Dynamic Compression for Novel Haptic Interactions

A Dissertation

SUBMITTED TO THE FACULTY OF THE

UNIVERSITY OF MINNESOTA

Wen Yen Esther Foo

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Advisor: Dr. Brad Holschuh

December 2020

Wen Yen Esther Foo

Acknowledgements

First and foremost, I would like to express my deepest gratitude to my advisor and Wearable Technology Lab (WTL) co-director, Dr. Brad Holschuh for his unwavering guidance throughout my time here at the University of Minnesota. He has given me the best experience a doctoral student could ask for, and I am very appreciative of him for entrusting me with the creative freedom to pursue ideas that interests me. I am also eternally grateful to our other lab co-director and committee member, Dr. Lucy Dunne for her valuable perspectives that has helped me grow tremendously as a researcher. I owe much of my professional and personal development to the mentoring and support they have both always given me—thank you both for all that you’ve done. I would also like to thank the other members of my advisory committee, Dr. Lana Yarosh and Dr. Peggy Martin, for their help, constructive feedback, and encouragement throughout this research process.

I am extremely thankful for the opportunity to be part of the WTL—I’ve met, worked with, and learned from many amazing people throughout my time here. I would like to express my sincere gratitude to my WTL co-authors on this project: Walter Lee, Simon Ozbek, Justin Baker, Crystal Compton, as well as undergrads: Miles Priebe, Mary Korlin-Downs, and Justin Barry, who have all contributed greatly to this research effort. Special appreciation for our WTL lab manager Heidi Woelfle for putting up with my never-ending requests and for being so patient in teaching me all that has to do with soft goods. Huge thank you also to all other WTL lab members—there are too many to name—for their positivity and support, and to our WTL alumni and my good friends Nika Gagliardi and Mary Ellen Berglund, for keeping me sane and motivated throughout the writing process (special thanks to Nika for some editing support as well). I am delighted to have worked with all of you and I wish you all the best in your future endeavors!

I owe a great deal of gratitude to both my parents who have been supporting me mentally and financially throughout my time here in the US. They never had the opportunity to attend college but have selflessly spent a significant portion of their lives making sure that I could enjoy all these opportunities. Without their support and sacrifice, I would not be here—I am forever i indebted to them. Finally and most importantly, my deepest appreciation goes to my husband, Ziqing Lin for his unparalleled love and support throughout this journey. Thank you for always supporting my crazy decisions of running around the country in the pursuit of my professional goals, for sharing your wisdom on how to navigate the academic world, and for all your encouragement that has motivated me till the very end. I know in my heart had I never met you, I would never have achieved this dream of mine. I am lucky to have you in my life and I am excited for what the future brings!

While I cannot comprehensively acknowledge all who have supported me throughout this journey, I am grateful for all the people who have been there for me, especially in the midst of a global pandemic. This has been a fantastic learning experience and a great journey that I will cherish for the rest of my life.

This work was supported by the National Science Foundation (Grant # 1656995), University of Minnesota Grand Challenges Research Grant, University of Minnesota Doctoral Dissertation Fellowship (DDF), and IEEE Robotics and Automation Society (RAS) Technical Committee on Haptics ‘Innovation in Haptics’ Program.

Abstract

The sense of touch is an integral part of our everyday experiences. One of the touch sensations that we experience ubiquitously is compression—common human interactions of being held, swaddled, hugged, and squeezed, all involve compressive forces on the body. The use of compression as an interaction modality offers advantages of resembling common human behaviors, and is capable of invoking a range of attention depending on compression features, while being less distracting than typical vibrotactile approaches. Compression as a form of haptic stimulus is also widely used in medical interventions (e.g., compression stockings/vests) and has the potential to be integrated into new research areas such as mediated social touch, distributed notification systems, and immersive experiences, yet we know very little about it. Currently, there is a lack of understanding of the perceptual/experiential impacts given varying compression parameters on large areas of the body, as well as the ability of extending this form of sensory stimulation to functional applications. Hence, this dissertation examines, in detail, on-body compression as a novel interface and mode of interaction: specifically, it seeks to answer questions related to how people experience compression stimulus given varying compression inputs, applications and contexts.

Motivated by the need to effectively study this problem, we also made strides in advancing compression technologies. Using a human-centered approach of gathering user feedback and incorporating them into design iterations, we developed soft, garment-based technologies capable of delivering dynamically controllable compression, involving varying upper-body locations, intensity, durations, and patterns. For the first time, we demonstrated the use of shape memory alloys (SMAs)—a type of soft robotic actuator—to controllably generate compression across distributed areas of the body. The use soft actuator elements integrated into garments are a crucial element in this development process; the garment platform offers intimacy, direct access to large body areas, social ubiquity, and enhanced wearability. Given the enabling technology (and also motivated by the interface/interaction opportunities afforded by this technology), we studied the parameters and confounds that influence user experience, taking

iii into consideration how the sensations were perceived, the effects on user comfort, and subjective preferences. The three major takeaways include: (1) importance of sizing and fit, (2) individual preferences and the need for customizability, and (3) the relationship between context-specific stimulation patterns and emotional reactions.

Based on the understanding of the experiential effects of on-body compression provided through the SMA-based garments, several candidate applications were down-selected and investigated. Given the close relationship between social touch and emotions, one area of huge potential involves technology-mediated affective applications (i.e., technology that support the detection, display, or communication of affect). Hence, leveraging the inherent SMA properties as being capable of providing warm, compressive actuation—acting as a proxy for human touch, we mapped the design space for garment-mediated emotional communication through warm, compressive forces. Two online surveys were deployed to gather user expectations in using garment-mediated warm, compression (of varying body location, intensity, pattern) to communicate 7 distinct emotions, while evaluating the range of mental models used. The findings show 5 major mental model categories: (1) representation of body sensations, (2) replication of typical social touch strategies, (3) metaphorical representation of emotions, (4) symbolic representation of physical actions, and (5) mimicry of objects or tasks. The employment frequency of each of these mental frameworks and haptic parameters were synthesized to inform future haptic garment design approaches for emotional communication. Further, from participant feedback and literature support that has provided evidence that compression could elicit positive affect or feelings of calm, another area of investigation was on the use of compression for affect modulation. Through a mindful meditation study augmented with compressive haptics delivered through the SMA-based garment, we demonstrate the potential of using compression to improve user’s meditation experience and in the long-run, help users more effectively regulate their emotions.

Ultimately, the results of this research give rise to new opportunities in a variety of applications and provide a roadmap for interface/interaction design in those context, including enabling new modes of interaction between users separated by distance (e.g., tele-rehabilitation, social mediated touch) as well as new haptic sensations in the area of immersive experiences (e.g., media augmentation, virtual reality).

Table of Contents

Acknowledgements ...... i

Abstract ...... iii

Table of Contents ...... v

List of Tables ...... ix

List of Figures ...... xi

1. Introduction...... 1

Research Context and Motivation ...... 3

Research Contributions ...... 5

Thesis Overview and Structure ...... 6

2. A Survey of Haptic Perception, Systems, and Applications ...... 8

Haptic Technologies and Applications ...... 8

Nature of Human Haptic Perception ...... 10

2.2.1. Sensory Receptors...... 11

2.2.2. Delineating the use of the term ‘Compression’ ...... 12

2.2.3. Psychophysics of Compression Perception ...... 14

2.2.4. Discriminatory and Affective Touch Dimensions ...... 16

Wearable Compression-Based Technology ...... 18

2.3.1. Information Display/Notifications ...... 19

2.3.2. Affect Communication and Social Presence ...... 20

2.3.3. Media Augmentation and Entertainment ...... 22

2.3.4. Persuasive Touch in Health ...... 23

Compression Technology Form Factors & Actuation Mechanisms ...... 25

Summary Table of Wearable Compression Technologies ...... 27

Research Gaps in the Study of Compression-based Wearables ...... 29

3. User-Centered, Iterative Design of Soft Robotic Actuated Compression Garment Technology ...... 34

Exploratory Study with Low Fidelity Prototype ...... 36

3.1.1. Low Fidelity Test Garment Design...... 36

3.1.2. Low Fidelity Prototype User Study Methodology ...... 38

3.1.3. Exploratory Study Results and Discussion ...... 39

3.1.4. Exploratory Study Insights and Design Implications ...... 43

Technology Investigation and Selection—Shape Memory Alloys ...... 44

3.2.1. Survey of Current Compression Technology Strategies ...... 44

3.2.2. Shape Memory Alloys (SMAs) ...... 47

Feasibility Study with Medium Fidelity Prototype ...... 54

3.3.1. Medium Fidelity Test Garment Design ...... 55

3.3.2. Medium Fidelity Prototype User Study Methodology ...... 57

3.3.3. Feasibility Study Results and Discussion ...... 59

3.3.4. Feasibility Study Insights and Design Implications ...... 62

High Fidelity Garment Prototype Design Iteration ...... 65

Technology Development Process- Summary ...... 70

4. Variables Influencing User Experiences of Computer-Mediated Compression ...... 73

Interaction Study Methodology ...... 74

Interaction Study Results and Discussion ...... 76

Interaction Study Discussion and Implications ...... 92

4.3.1. Limitations and Future Work ...... 95

Interaction Study Summary ...... 96

5. Compressive Haptics in Emotional Communication ...... 98

Related Work ...... 99

Survey Study 1: Gathering ‘Warm Touch’ Strategies and Mental Models Employed in Emotional Communication ...... 102

5.2.1. Methods ...... 102

5.2.2. Results: Quantitative Survey Data ...... 105

5.2.3. Results: Qualitative Open-Ended Responses ...... 108

5.2.4. Survey 1: Summary ...... 113

Survey 2: Investigating If Communication Objective Changes User’s Mental Models In ‘Warm Touch’ Haptic Communication ...... 114

5.3.1. Methods ...... 114

5.3.2. Results: Quantitative Survey Data ...... 115

5.3.3. Results: Qualitative Open-Ended Responses ...... 119

5.3.4. Survey 2: Summary ...... 122

Discussion and Implications ...... 123

5.4.1. Limitations and Future Work ...... 127

User Expectations of Garment-Mediated Emotion Communication- Summary .... 128

6. Compressive Haptics for Emotional Modulation ...... 130

Related Work ...... 131

Compression-Actuated Haptic Garment Design ...... 134

Study Design and Data Collection ...... 138

Results and Discussion ...... 140

Compression-Augmented Mindfulness Study Summary ...... 145

6.5.1. Discussion and Limitations ...... 145

6.5.2. Broader Implications for Persuasive Technology in Emotion Regulation ... 148

7. Conclusion ...... 150

vii

Summary of Key Findings ...... 150

Limitations and Future Directions ...... 153

Applications and Broader Implications ...... 156

7.3.1. Affective Communication ...... 156

7.3.2. Media Augmentation and Multimodal Interactions ...... 157

7.3.3. Information Transfer and Ambient Notifications ...... 157

7.3.4. Persuasive Technology ...... 158

7.3.5. Dynamic Garments as Expressive Elements ...... 158

7.3.6. Ethical Considerations ...... 159

Final Remarks ...... 160

List of Associated Scholarly Work ...... 160

Bibliography ...... 163

Appendices ...... 187

High-fidelity Garment Perceived Intensity Statistical Results Summary Table ..... 187

High-fidelity Garment CALM Pressure Comfort Statistical Results Summary Table ...... 188

Emotion Communication Survey Study Participants’ Demographic Breakdown .. 190

Emotion Communication Survey 1 Study’s Parallel Sets Plot Data Visualization 191

Emotion Communication Survey 1’s Pairwise Comparisons of Confidence Ratings between Emotions ...... 193

Emotion Communication Survey 2 Study’s Parallel Sets Plot Data Visualization 194

Emotion Communication Study Proposal: Encoding-Decoding Communication Task Mediated by SMA Compression Garment ...... 197

Haptic-Augmented Meditation Study’s Force Sensor Calibration and Measurement ...... 208

Emotion Regulation Study Proposal: Stressor Task SMA-based Garment Evaluation ...... 210

viii

List of Tables

Table 2-1: Summarized list of compression-based wearables, categorized based on the most common wearable haptic actuation technologies: (a) servo motor-based solutions, (b) pneumatically-driven inflatable solutions, and (c) shape memory alloy solutions . 27

Table 3-1: Evaluation of common haptic actuator types and their associated advantages and disadvantages as potential compressive actuation strategies...... 45

Table 3-2: Variable definitions for predicting active pressure applied through SMA-integrated garment ...... 51

Table 3-3: Low to high fidelity garment iterations [167] ...... 72

Table 4-1: Scenario association in which situations participants expect the garment to be used are summarized and presented ...... 89

Table 5-1: Prompts for Emotional Communication Task ...... 104

Table 5-2: Survey 1 participants’ mental model strategies used to communicate emotions .. 109

Table 5-3: Alternative haptic strategies proposed by Survey 1 participants ...... 112

Table 5-4: Survey 2’s Fisher's Exact Test for comparison between groups ...... 116

Table 5-5: Survey 2 Group 1’s (Communicate) mental models used during‘warm touch’ strategy selection ...... 120

Table 5-6: Survey 2 Group 2’s (Elicit) mental models used during‘warm touch’ strategy selection ...... 120

Table 5-7: Alternative Haptic Strategies Proposed by Users Divided by Group ...... 122

Table 5-8: Possible design strategy for Affective Garment to be used for Communication and Elicitation Contexts ...... 126

Table 6-1: Selected Flow State Scale (FSS) items ...... 145

Table B-2: CALM Pressure comfort ANOVA summary table, comparing conditions between various body locations ...... 188

Table B-3: Tukey multiple comparisons of means summary table for the upper arm...... 189

Table B-4: Tukey multiple comparisons of means summary table for the shoulders ...... 189

Table C-5: Country of origin information for participants of both surveys ...... 190

Table C-6: Number of years participants from both surveys have spent in the United States 190

Table E-7: Pairwise comparisons of confidence ratings using paired Wilcoxon rank sum test with Bonferroni corrected p-values...... 193

Table H-8: Analog voltage readings collected by Data Acquisition System that are used to calculate estimated pressures provided by the garment ...... 209

List of Figures

Figure 2-1: Comparison of two-point discrimination for touch as measured by Weinstein (1968), Weber (1834), and Mancini et al. (2014) [88]–[90]. Figure obtained from Mancini et al. [90]...... 15

Figure 3-1: Low-fidelity garment- Front view (Left), Side view (Middle), Back View (Right). Hook-and-loop straps are anchored strategically on varying garment locations to allow for maximal adjustability and flexible application of compression on various body regions ...... 37

Figure 3-2: Shoulder test conditions with changing compression vectors: (a) ‘Straight’, (b) ‘One- side crossed’, (c) ‘Diagonal crossed’, and (d) ‘Mixed’ ...... 38

Figure 3-3: Frequency count of participants’ perception of compression distribution for various compression locations and vectors for low fidelity garment ...... 40

Figure 3-4: Distribution of participants’ shoulder compression vector preferences for low fidelity garment ...... 40

Figure 3-5: (Left) Method of producing tightly packed coils from SMA wire. (Right) Comparison of tightly packed coils vs. SMA coil that has been physically stretched out. Obtained from [160] ...... 48

Figure 3-6: SMA coil actuator shape setting steps and activation cycle. Figure obtained from [160], [161] ...... 49

Figure 3-7: Simplified compression system activation process, with SMA and passive fabric coupling, as well as representative force-length relationship for the described SMA- based compression system. Figure adapted from [123]...... 50

Figure 3-8: Representative illustration of the SMA-actuated compression system wrapped xi

circumferentially around a body part ...... 51

Figure 3-9: Prototype iterations of SMA-integrated garments. (a) Left: SMA connected in series; (b) Right: SMAs in parallel in conjunction with softer fabrics ...... 56

Figure 3-10: Final SMA-based medium-fidelity garment prototype’s illustration and photos. The garment consists of an inner comfort layer (beige) and outer actuation layer (gray) with SMA actuators connected in a parallel circuit configuration...... 56

Figure 3-11: SMAs integrated into the medium fidelity prototype garment—screenshots gathered from a video showing SMA activation from a relaxed to fully contracted position on a mannequin shoulder. (Left):Loose garment, SMAs are relaxed; (Right): Upon electrical activation, SMAs contract into tightly packed coils and compression is applied ...... 57

Figure 3-12: Medium fidelity garment shoulder test conditions (a) ‘Straight’ (b) ‘Diagonal’, (c) ‘Mixed’ shoulder compression vectors ...... 58

Figure 3-13: Participants’ shoulder actuation vector preference for medium fidelity prototype ...... 59

Figure 3-14: High fidelity, garment-based dynamic compression system components and design. (A) Inner comfort layer and middle actuation layers with integrated SMA actuators on the torso and shoulders; (B) Final men’s and women’s garment including outer covering and arm bands ...... 65

Figure 3-15: SMA-braid actuators integrated into garment—screenshots gathered from video showing SMA activation from a relaxed to fully contracted position on a mannequin torso. (Left) Unpowered, relaxed garment state. (Right) Powered, compressed garment state. When unpowered, the SMAs will slowly relax and return to a loose garment state (Left) due to the elastic energy stored in the braided sheath...... 66

Figure 3-16: Force output by a single SMA-braid actuator measured with an Instron tensile testing machine. The SMA-braid structure was powered on for approximately 40 seconds, and power was shut off for the remaining of the test capture. The compression-relaxation behavior and force increments/decremetns during power xii

on- and off- periods can be observed...... 67

Figure 3-17: SMA actuator channel distributions. The numbering of SMA actuators were based on their position on the garment: starting from crainal to caudal for the torso and arms, and lateral to medial for the shoudler actuators. For low intensity compression, only channel A was activated; for medium intensity compression, channels A + B were activated; for high intensity compression, all channels A + B + C were activated. Note the additional recruitment of actuators with the addition of channels, resulting in increasing compression levels ...... 68

Figure 3-18: Three compression intensities visualized via a thermal camera – greater pressures are generated by recruiting additional parallel actuators (since the SMAs activate using Joule heating, actuator activation visuals can be captured by a thermal camera) ...... 68

Figure 3-19: Processing user interface for the garment compression control. Selection of buttons on this computer interface sends a signal to the garment’s on-board electronics via Bluetooth; a variety of compression parameters can be controlled remotely ...... 69

Figure 3-20: Representative pressure measurement map showing relative pressure distributions on a mannequin side provided by the high-fidelity SMA garment. (A) no compression activation—only contact pressure by the garment (i.e., passive pressure), (B) low intensity acutation, (C) medium intensity actuation, and (D) high intensity actuation (i.e., active pressures)...... 70

Figure 4-1: The high fidelity SMA-integrated garment worn by a female. SMA actuators with controllable intensity levels were located on the torso, arms, and shoulders regions...... 74

Figure 4-2: Participants were asked to rate stimuli perception in each of these numbered body regions to prevent confusion between described body regions. This also acts as key/legend for the figures below...... 76

Figure 4-3: Perceived compression intensity ratings for Torso (sides, abdomen, lower back) and Arms. The lower and upper box boundaries are of the 25th and 75th percentiles; the

xiii

center line represents the median; the center cross marker indicates the mean; the whiskers are of the 10th and 90th percentiles. (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’) ...... 77

Figure 4-4: Perceived intensity ratings for ‘Shoulder-Straight’ compression vector. The upper torso regions likely affected by this compresison vector include front side chest, front mid chest, top of shoulderes, back of neck, and upper neck. The lower and upper box boundaries are of the 25th and 75th percentiles; the center line represents the median; the center cross marker indicates the mean; the whiskers are of the 10th and 90th percentiles. (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’) ... 79

Figure 4-5: Perceived intensity ratings for ‘Shoulder-Diagonal’ compression vector. The upper torso regions likely affected by this compresison vector include front side chest, front mid chest, top of shoulderes, back of neck, and upper neck. The lower and upper box boundaries are of the 25th and 75th percentiles; the center line represents the median; the center cross marker indicates the mean; the whiskers are of the 10th and 90th percentiles. (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’) ... 79

Figure 4-6: Compression comfort ratings of torso regions (trunk of body, abdomen, middle/lower back) and Arms. Note gender-separated averaged results (Triangle markers= males; Circle markers= females) with unidirectional standard deviations bars (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’) ...... 81

Figure 4-7: Compression comfort ratings of upper torso & shoulders. Note gender-separated averaged resultswith unidirectional standard deviations bars (Triangle markers= males; Circle markers= females) and shoulder compression vectors (vertical marker lines= ‘Shoulder Straight’; diagonal marker lines= ‘Shoulder diagonal’. (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’) ...... 82

Figure 4-8: Liking/Disliking of garment compression rated on a 7-point scale. The frequency of participants for each category are indicated in the figures...... 83

Figure 4-9: (Left (a)) Compression stimulus presentation preference; (Right (b)) Torso compression stimulus intensity preference by gender and personal preference for hugs ...... 84 xiv

Figure 4-10:Liking/Disliking of garment temperature rated on a 7-point scale. The frequency of participants for each category are indicated in the figures...... 85

Figure 4-11: SAM Valence ratings for all test conditions. The lower and upper box boundaries are of the 25th and 75th percentiles; the bolded center line represents the median; the whiskers are of the 10th and 90th percentiles; outlier data falling outside the lower and upper quartile ranges are indicated as circles...... 87

Figure 4-12: SAM Arousal ratings for all test conditions. The lower and upper box boundaries are of the 25th and 75th percentiles; the bolded center line represents the median; the whiskers are of the 10th and 90th percentiles; outlier data falling outside the lower and upper quartile ranges are indicated as circles...... 87

Figure 4-13: SAM Dominance ratings for all test conditions. The lower and upper box boundaries are of the 25th and 75th percentiles; the bolded center line represents the median; the whiskers are of the 10th and 90th percentiles; outlier data falling outside the lower and upper quartile ranges are indicated as circles...... 87

Figure 4-14: Free Word Association Word Cloud ...... 89

Figure 4-15: Physiological data case-study with two participants; P15 and P13. The red vertical lines dividing the physiological data divides the different test conditions experienced by participants; the heart marker indicates each participant’s self- reported favorite compresison parameter. Note that P15 did not indicate a preference for arm compressions...... 91

Figure 5-1: Representative survey interface presented to users. (a) Top: survey interface that records participants’ ‘warm touch’ strategies corresponding to each emotion; (b) Bottom: reference chart with parameter definitions ...... 103

Figure 5-2: Representative parallel sets plot for the emotion Fear. This data visualization strategy was selected to demonstrate the relationship between parameter dimensions. For each ‘warm touch’ parameter dimension (body location, intensity, pattern), a vertical bar is shown for each possible variable; the length of the bar represents the frequency of user selection. ‘Ribbons’ connect each parameter dimension, showing

how the categories are distributed and related ...... 105

Figure 5-3: Survey 2’s Pearson residuals plot for Top: Body Location vs. emotions, Middle: Intensity vs. emotions, Bottom: Pattern vs. emotions. The strength (size of circle) and direction (positive=blue and negative=red) of associations between each parameter and emotion are represented...... 106

Figure 5-4: Survey 1’s balloon plot representing frequency of selection of all possible body location combinations and ‘No Touch’ selections. (Note that from top to bottom, the number of body locations involved increases) ...... 108

Figure 5-5: Survey 2’s Pearson Residuals Plot for Row 1: Body Location vs. Emotions, Row 2: Intensity vs. Emotions, Row 3: Pattern vs. Emotions for Group 1- Communicate (Column 1) and Group 2- Elicit (Column 2). The strength (size of circle) and direction (positive=blue and negative= red) associations are visually presented 116

Figure 5-6: Parallel Set Plots for Happiness. Left: Group 1 (Communicate); Right: Group 2 (Elicit)...... 117

Figure 5-7: Parallel Set Plots for Fear. Left: Group 1 (Communicate); Right: Group 2 (Elicit) ...... 117

Figure 5-8: Balloon Plot of Selection Frequency for All Possible ‘Warm Touch’ Combinations. Left: Group 1- Communicate; Right: Group 2- Elicit ...... 118

Figure 6-1: A typical mindful awareness practice session—Mind wandering is a typical occurrence and happens in a cyclical, iterative process...... 133

Figure 6-2: Men (Left) and Women's (Right) compression garment prototype with demonstrated garment (comfort/under) layer and SMA compression actuators located on the shoulder regions...... 135

Figure 6-3: SMA actuators unpowered/relaxed (Left) vs. powered/ compressed (Right). Cyclic compressions are provided with the compression-relaxation behavior of the SMAs housed within the braided sheath...... 136

xvi

Figure 6-4: Positioning of breath rate sensors on the diaphragm and the abdomen for redundancy and to accommodate possible differences in breathing styles. Center (black) portion of the sensors contains the coverstiched traces whereby changes in resistance were recorded given torso volumetric changes as the participant breathes...... 137

Figure 6-5: Compression haptics-guided meditation user study procedure. This employs a within-subjects study design, in which each participant experiences both the with- and without- garment conditions during the 1st or 2nd meditation task (the order of which is randomized)...... 138

Figure 6-6: A representative graph (P1) collected from BR sensor. Note the prominent peaks corresponding to each breath...... 139

Figure 6-7: Participant responses in comparing sensations provided by the SMA-based compression garment to everyday activities...... 141

Figure 6-8: Differences in breath rates between control and intervention; results were rank ordered for the x-axis. Participants P3, P4, P5, P8 did not prefer the haptic garment...... 142

Figure 6-9: Differences in heart rates between control and intervention; results were rank ordered for the x-axis. Participants P3, P4, P5, P8 did not prefer the haptic garment...... 142

Figure 6-10: Differences in the number of SCRs between control and intervention; results were rank ordered for the x-axis. Participants P3, P4, P5, P8 did not prefer the haptic garment...... 143

Figure 6-11: Differences in the number of instances of loss of focus between control and intervention; results were rank ordered for the x-axis. Participants P3, P4, P5, P8 did not prefer the haptic garment...... 143

Figure 6-12: Semantic Differential Scale (SDS) results. Differences in ratings were observed between the control (no garment) and intervention (with garment) conditions for positive-negative, active-passive, warm-cold scales...... 144

xvii

Figure 7-1: Human-machine system. Image obtained from [216] ...... 151

Figure D-2: Survey 1’s parallel sets plots data visualization for Ekman’s basic emotions. (a) Row 1 left: Sadness, (b) Row 1 right: Happiness, (c) Row 2 left: Fear, (d) Row 2 right: Anger ...... 191

Figure D-3: Survey 1’s parallel sets plots data visualization for prosocial emotions and other intentions. (a) Row 1 left: Love, (b) Row 1 right: Gratitude, (c) Row 2 left: Calm, (d) Row 2 right: Attention ...... 192

Figure F-4: Survey 2’s quantitative survey data visualization for Ekman’s basic emotions First column: Group 1 (Communicate); Second column: Group 2 (Elicit) ...... 195

Figure F-5: Survey 2’s quantitative survey data visualization for prosocial and other emotions. First column: Group 1 (Communicate); Second column: Group 2 (Elicit) ...... 196

Figure G-6: Prototype SMA-based warm touch’ garments designed based on the insights gathered from this survey study. The garments will be used to investigate if and/or how emotional communication and elicitation can be mediated by such wearable technologies [218] ...... 198

Figure H-7: Fabric-based force sensors (two conductive materials sandwiching Eeyonyx material) manufactured according to [226] ...... 208

Figure H-8: Force Sensor Calibration Curve ...... 209

Figure H-9: Haptic-driven emotion regulation (neutral-high arousal, with stressor tasks) study design overview ...... 210

xviii

Chapter 1

1. Introduction

Touch is the first of our senses to develop and is an important way for us to perceive and represent reality. We receive touch information from the world around us every minute and every second of the day—from pressure on the soles of our feet from walking, to the shear forces when we swipe our fingertip across the screen of our phones—the full experience of all these simple actions would not be possible without the sense of touch. It is so central and omnipresent in our lives that it would be quite impossible to imagine our lives devoid of it. The sense of touch is commonly associated with sensory-discriminative functions, involving kinesthetic and cutaneous mechanisms that encode stimulus characteristics (e.g., intensity and spatial localization), allowing us to obtain information about our environment [1], [2]. For instance, we use touch to identify object features given the perception of pressure, texture, slip, et cetera, on our hands and fingers. However, apart from these discriminative purposes, the human sense of touch also serves an affective purpose, playing a role in how we interact with others [1], [3]. Unlike most of the other senses (e.g., visual, auditory, olfactory), touch interactions always occurs within one’s peripersonal space (i.e., the space immediately surrounding our bodies), where objects and people are within reach [4], [5]. When touch occurs between two or more individuals, it typically implies an interaction with another person, and this is known as interpersonal touch [6].

Interpersonal touch is extremely important in human development; it plays a critical role from infancy and throughout an individual’s lifespan. From an early development point of view, one of the most classic experiments was conducted by Harry F. Harlow in 1958 [7]. Harlow allowed for the infant monkeys to either be ‘raised’ by either a terry cloth or wire mesh surrogate mother. These two groups of surrogates were then either divided to provide milk or otherwise. Surprisingly, Harlow found that the infant monkeys preferred the cloth surrogate without milk, over the wire surrogate with the milk, suggesting that touch at that developmental stage is as much as, or if not more, important than nourishment. Further, as the monkeys grew, the infants exposed to the wire mesh surrogate appeared to develop abnormal grooming patterns and even exhibited pathological violent tendencies [7], [8]. These effects were also seen in cases of children who were touch deprived, exemplified by the tragic case of a group of Romanian orphans found in overcrowded institutions that were barely touched by their caretakers due to the lack of resources [9]. The consequences were dramatic; these children didn’t just grow up to have a host of emotional problems (e.g., depression, violent tendencies, schizophrenia), but they also had a multitude of physical illnesses including suppressed immune system, delays in growth, and skin ailments [6], [9], showing the central role touch plays for both physical and cognitive development.

However, interpersonal touch isn’t only vital in our early developmental years, but its significance persists throughout the rest of our lives—humans are inherently social creatures and touch is a crucial facet to social interactions [10]. Whether it is a firm handshake, a pat on the shoulder, a grasp for attention, or a peck on the cheek, the sense of touch provides us a mode of communication. It can be used convey affective information [5], to elicit emotional experiences in others [5], [11], or enhance the meaning of other forms of verbal or non-verbal communications [11]. Touch has a powerful affective component that is reflected even in our basic biology, where there is mounting evidence that affective touch has a distinct somatosensory dimension compared to discriminative touch [1]. Beyond affective communication, touch also plays a role in regulating and influencing broader social interactions [12]; it could be said to be a form of ‘social glue’. Our personal experiences can speak to these ideas as well. The types of touch are wide-ranging— from greetings, to showing gratitude and trust—touch is used in connecting people in the community, reinforcing familial bonding, fostering romantic or sexual relationships, and even governing our own emotional well-being [6], [13], [14].

It is undeniable how important the sense of touch is in our everyday lives, yet it is one of the most

2 overlooked given the historical lack of interest in formally studying touch compared to the other senses (e.g., vision and audition). However, in the recent years, researchers have begun to appreciate the richness of this sensory modality; it is now being pursued from all directions—from biology, including cognitive and neural correlates [1], [11], to novel technologies in a variety of application areas, including new communication paradigms, interactive systems, medical interventions, and emotional modulation, just to name a few [5], [10], [15]. The pursuit of scientific knowledge regarding the sense of touch is known as the field of haptics.

Research Context and Motivation

The term haptics can be traced back to its Greek origins, with the Greek verb haptesthai taken to mean ‘to touch’ or ‘to contact’ [15]. Currently, the term has been adopted by many different disciplines of studies and encompasses not only the active touch of objects by humans, but also all aspects of human-computer touch interactions or machine touch [15]. Touch can be defined as a composite of a variety of sensations, through the stimulation of the skin by mechanical, thermal, chemical or electrical stimuli. Given the myriad uses of touch in our everyday lives, it comes as no surprise that a great deal of effort is dedicated to developing artificial systems, also known as haptic devices, to capture, produce, and study the salient features and nuances of human touch [16]. With that, an assortment of sensations can be elicited, such as temperature, vibration, pressure, tapping, and compression. Interestingly, in stark contrast to the diversity of touch sensations in our everyday lives, the use of haptics in current consumer market products are largely limited to only vibrotactile feedback (e.g., cell phone notifications or feedback in game consoles), but there is so much more to be explored in this regard. Many natural human touch sensations take the form of sustained or distributed forces on the body, which is not in line with the type of haptic deployment we see in consumer devices. In fact, even in the real world, very few everyday experiences are conveyed through vibration alone [17]. More sustained and distributed forces could give rise to varied sensations, colloquially referred to as pressure, squeeze, or compression (the distinction of these terms will be discussed in Chapter 2). The study of such forces is still relatively unexplored in the field of haptics, and therefore, this research focuses on one of these understudied haptic modalities—compression.

Compression is experienced in a plethora of contexts, from the everyday ubiquity (e.g., feeling pressure on the soles of our feet or wearing form fitted clothing), to the clinical (e.g., graduated

3 stockings to improve circulation), to the emotional (e.g., receiving a hug from a loved one), to the pleasurable (e.g., receiving a therapeutic massage) and the social (e.g., shaking hands with a new neighbor). Given that we as humans experience compression so regularly in our everyday lives, it is perhaps a little surprising how—with only a handful of exceptions—it is almost completely overlooked as a haptic modality or interface. Compressive-based forces are particularly interesting because it is a type of sensation that is already known to be similar to common human behaviors and actions. The ubiquity of the sensation could give rise to perceptions of being more ‘natural’ or ‘pleasant’ than other forms of haptic stimulation (e.g., vibrations) [18], [19]. In fact, in the field of occupational therapy, compression is used as a clinical intervention in deep touch pressure therapy, whereby strategically-applied compression is used to improve the general well-being of users by decreasing anxiety, promoting feelings of calmness and improving alertness [9], [20]–[22]. This could mean that compression might offer potential in modulating a user’s emotional state. In the technology realm, user interactions with various information types could be augmented with multimodal display technologies, yet compression as a display modality is currently underexplored [23], [24]. Besides, the sense of touch is known to carry huge emotional significance [25]. Hence, compression, being a sensation that is comparable to many human interactions, may offer rich, visceral experiences. The affective components associated with compression could be translated given new technologies, driving the opportunity for rich affective experiences, such as augmenting the experience of media or immersive environments (e.g., augmented reality (AR)/ virtual reality (VR)) [17], [26], [27], communicating emotional messages [28], [29], and even creating persuasive experiences by leveraging emotional responses to compression sensations [30].

Haptic technologies can take a variety of form factors, from earth-grounded devices [31], [32], robots or dolls [33], to wearable systems—ranging from wristlets [34], [35], gloves [36], sleeves [37], [38], to full-body garments like jackets and vests [27], [39], [40]. The focus of this research will be on body-scale garments since they afford interactions with multiple body regions (more similar to natural touch behaviors as opposed to localized systems limited to only a small body region, such as the wrists/arm), while allowing the display of more haptic information given large, distributed areas. In addition, the garment platform is ideal for generating on-body compression since it is intimate, has close proximity to the body, offers direct access to large areas, and is socially ubiquitous. Further, the ‘soft’ garment structure (given the utilization of fabrics) that theoretically features a smaller mechanical impedance differential between the haptic device and a user (as compared to traditional robotic systems such as hard exoskeletons), should provide better fit,

4 usability, and wearability—all of which are important dimensions contributing towards a user’s overall, positive experience with a wearable haptic system [38], [41].

Given the ubiquity of compression in our everyday lives and potential use of compression in these wide fields of applications, there are many open questions to be answered before we can fully understand (and leverage) the effects of compression stimulation on the body. The experience of compression can be modulated with varying inputs (e.g., location, intensity, rate, duration, pattern), but not much is known in regard to proper stimulation parameters for a desired experience. In addition to understanding how perception of the varying compression inputs relate to a user’s subjective experience, the characterization of the relationship between the aforementioned compression parameters and their potential affective effects (e.g., for communicating emotions or modulating affect) is also necessary given their use potential in various application spaces. Hence, this dissertation investigates how people experience compression stimuli on the body given varying compression inputs, applications and contexts—towards the goal of body-scale devices (as opposed to localized systems on small body areas like the wrist)—by using soft garment form factors to deliver compression stimuli.

Research Contributions

This work first investigates user perception of on-body compression, broadly, from an experience standpoint (including the consideration of parameters/confounds that influence user experience), and later examines application-specific use cases to investigate the use of compression from a functional standpoint, with particular focus on its use in communicating and modulating emotions. In order to understand, evaluate, and deploy garment-based compression, however, advancements in wearable compression technology are also required. Therefore a portion of this work is dedicated to the design and development of novel compression garment technologies—given recent advances in soft robotic actuation technologies—that ultimately enable new types of on-body haptic (compression) experiences. The contributions of this work include:

1. Soft Robotic Actuated Garment Technology Development Design and development of soft, wearable upper body compression garments integrated with soft robotic actuation technologies that are capable of delivering controllable dynamic compression inputs (location, intensity, duration, pattern).

2. Identification of User Perceptions and Preferences for Compression-based Haptic Systems Assessment of the subjective, experiential effects of garment-based, computer-mediated compression, including user perception and preferences, given varying compression stimuli parameters on several upper body locations.

3. Mapping of the Design Space for Compression-mediated Emotional Communication Characterization of user expectations and mental models associated with computer-mediated compression as a communication and display modality, as well as the development of design strategies for future haptic communication applications utilizing soft robotic-actuated compression garment technologies.

4. Determination of the Potential of Compression Systems in Affect Modulation Applications Demonstration of the application potential of compression-based interactive systems from an affective (functional) standpoint—given a compression-augmented mediation scenario.

Thesis Overview and Structure

This thesis is divided in to 7 chapters. There are two major parts in this dissertation. The first group of chapters (2-4) presents a series of investigations into technologies to enable garment-based on- body compression, the associated variables and parameters of on-body compression, and their impacts on user experience; the second group of chapters (5-6) presents user-driven design strategy development and user testing for targeted applications.

Chapter 2: An overarching review of the literature on compression-based systems (with an emphasis on those with soft, wearable garment form factors) was conducted to understand the research landscape and identify gaps that motivated the research.

Chapter 3: This chapter focuses on the end-to-end design process of soft robotic actuated compression garments while characterizing participant responses to the novel form of haptic stimulation. Taking a human-centered design approach, several compression garment iterations (low to high fidelity prototypes with varying degrees of technology integration) were developed and the resulting high-fidelity garment will be used as a research tool that enables the experiential and functional evaluation of on-body compression. The design space for dynamic 6

compression technologies is mapped and presented alongside design lessons learned throughout the process.

Chapter 4: Given the soft robotics-actuated garment innovation derived from Chapter 3, an investigation into the set of compression inputs, including location, duration, intensity, and pattern, was conducted and the impacts of these parameters and associated confounds have on user experience is clarified.

Based on user feedback from Chapters 3-4, several candidate application areas are down selected and further investigated functionally to demonstrate the potential of the system technology in various application spaces.

Chapter 5: This investigation involves characterizing the design space for emotional communication through warm, compressive actuation. Through two online user studies, we captured user expectations and perceptions in using this form of haptic strategy to communicate emotions, while evaluating and delimiting the range of mental models used by participants during the communication process.

Chapter 6: This chapter demonstrates the application potential of SMA-based compressive interactive systems from a functional standpoint—given a compression-guided mediation scenario. Specifically, we investigated the use of the SMA-based compression garment as an augmentation tool to a mindful-meditation practice with the aim of promoting focused attention in novice practitioners and an overall positive, meditation experience.

Chapter 7: A conclusion and summary of the thesis is presented, including major findings, limitations, broader impacts of the technology, and opportunities for the future.

Chapter 2

2. A Survey of Haptic Perception, Systems, and Applications

This investigation sits at the interface of several disciplines including human-computer interaction, haptics, wearable technology, human factors, and user experience. To understand the context for, and broader impacts of this investigation, this chapter begins with an overview of the larger field of haptic technologies, followed by the biology of human haptic perception, before examining past research in wearable compression-based haptic devices, and ending with a discussion on where this work sits within the domain of on-body haptics.

Haptic Technologies and Applications

The need for the sense of touch in our everyday lives is irrefutable—we use it to explore our environment, manipulate objects and communicate with others. Therefore, it comes as no surprise that a great deal of effort is dedicated to developing artificial systems, also known as haptic devices, to capture and produce the salient features and nuances of human touch [31]. When a touch interaction occurs, there are two facets to the interaction: our human haptic sense and the haptic

8 interface. Our haptic sense can be broadly classified into two modules: (1) cutaneous sensations (i.e., sensations gained from stimulus to the skin) and (2) kinesthetic sense (i.e., internal signals from tendons and muscles that provides knowledge about position, forces, and motion of a limb). On the other hand, haptic interfaces can be defined as ‘anything a user touches or is touched by, to control, experience, or receive information from’ [16]. Most of the research in this area are driven by haptic devices, mechanical apparatus used to exchange forces between a user and a computer. From this technological standpoint, haptic interfaces are devices that display tactile feedback (rendering a percept of the cutaneous sense—e.g., vibration, texture, compression, temperature) or force feedback (giving a user information on an object’s weight, hardness and inertia) to a user in the course of an interaction [42], [43].

Force feedback devices are usually directed force-displacement relationships that provide counter- forces that react to our movement in space [44]. These devices typically have multiple degrees of freedom (DoF), with significant research effort dedicated to direct rendering of sensations and ensuring stability and realism of virtual environments [45], [46]. Traditionally, these devices are earth-grounded (e.g., attached to a table) but may also feature other graspable systems (e.g., hand- held tools), or even body-grounded wearable devices (e.g., exoskeletons) [44], [47]. The ‘grounding’ of these devices allows the translation of forces and torques about a human joint, allowing the users to push on them (and be pushed back), hence, felt as movement or resistance. Force feedback devices are used in a variety of applications including training simulators, tele- operation of robotic devices (e.g., surgeon performing robotic assistive surgery, remote operation of a repair robot by an astronaut) [48], [49], or experiencing realistic physical interactions in VR environments [50]–[52].

Cutaneous-based tactile devices are different in that they are not concerned with forces across a user’s joint, but rather on stimulating the skin. It should be noted that tactile devices are commonly associated with vibration feedback, but here, we employ the term ‘tactile devices’ in a much broader sense—devices that provide cutaneous stimulation for tangible perceptions of touch, which may include anything from taps, vibration, temperature, point pressure and even distributed pressure. Hence, they are well suited for radically different contributions, perhaps most significantly as a communication modality or information display. Given the proliferation of electronic technology, the visual and auditory systems have been shouldering significant attentional demands for conscious interactions. As such, researchers are turning to haptics as a secondary form of

9 communication. The use of haptics for information display has roots in technologies to aid the visually impaired as a guidance tool for navigation [53], [54], and later expanded to be included in everyday mobile devices. It can be used as notification systems to notify changes in state, usually aiming to maintain a relatively non-intrusive ambient background awareness such that a user is capable of multitasking [45]. Because the tactile modality is a less common use for displaying information, it presents a cognitive advantage; by employing a different sensory system, tactile displays can be effectively used to supplement other modalities (e.g., vision, auditory), in conditions where they are overloaded [55], [56]. For instance, for a pilot flying a plane that involves high physical and visual demand, additional information presented (concurrently with the visual and physical tasks) through the tactile modality may be more easily understood than presenting all the information to the most occupied modality—vision.

Tactile devices can also be used to augment media experiences as a means to enhance an experience or add realism to interactions. It is perhaps most commonly explored as a means to capture the feel of interactions, simulating textured surfaces [57] or illusory sensations such as motion [58], [59]. It is used in a variety of applications including enriching experiences in storytelling [60], [61], entertainment/gaming [62], [63], educational media [64], or virtual stimulations [26], [65], [66]. But perhaps what really makes cutaneous tactile devices stand out in the world of haptic applications is the affective dimension that it could bring. Haptic elements can not only be used to enhance the affective components in an interactive experience [67], but it could also serve to deliberately convey emotions to another person [16], [31], [68], or even create persuasive experiences by leveraging emotional responses to haptic sensations [22], [30], [69], [70]—all of which are driven by the unique ability of the sense of touch in eliciting and modulating human emotion [11]. An astonishing variety of applications can arise from this concept of affect communication, prominently highlighted by the field of mediated social touch, whereby people can use their sense of touch in interactions with others through technology [5], [10]—from enriching communication channels such as messaging systems to increase connectedness [71]–[74] to collaborative tasks [75], the potential for using haptics in affect communication is limitless.

Nature of Human Haptic Perception

Before diving deeper into the discussion on the haptic interface modality that is the focus of this research work, we first take a brief detour to understand the biology that makes up our haptic sense.

2.2.1. Sensory Receptors

Our haptic sense is driven by the somatosensory system. It provides information about the environment to an individual and is involved in the conscious perception of touch, including pain, pressure, temperature, movement, vibration and position. Most of the body’s sense of touch occurs through the skin; the skin senses a wide variety of impulses through specialized receptors that are involved in sensing, translating mechanical stimuli into electrical signals and relaying information along nerve cells to the spinal cord and/or central nervous system [76].

The skin is a fascinating biological system—it forms the basis of all our physical interactions with the external world and it is bristling with sensors of various types and is largely governed by the somatosensory system. Arguably, the structures that are most pertinent to our investigation are sensory receptors; the processing of external stimuli is initiated by the activation of populations of sensory receptors at the body surface, before information is relayed to the brain for interpretation and action. Sensory receptors are specialized structures that detect changes in the environment, and they can be classified as either mechanoreceptors (detects light touch, pressure, and vibration), thermoreceptors (detects temperature), nociceptors (detects pain or itching), or proprioceptors (detects position and movement) [76], [77]. Since the overarching goal of this research topic centers on compression, we will be focusing on cutaneous mechanoreceptors which include Meissner’s corpuscle, Pacinian corpuscle, Merkel’s disc, Ruffini corpuscle [78]. Each of these mechanoreceptors respond to a specific type of stimulus by transforming the stimulus into electrical signals and transporting them to the brain. They can be broadly classified based on their temporal resolution (i.e., how quickly they adapt) and the size of their receptive fields (i.e., providing information about precision, and sensitivity). When a mechanoreceptor receives a stimulus, action potentials are fired at an elevated frequency. The rate of adaptation depends on the time it takes the action potential generated by the receptor to decay and ‘adapt’ to a constant stimulus (i.e., electrical signal pulses returning to its resting state). Rapidly adapting receptors respond maximally to stimuli, but this respond is brief as their signals decrease as the stimulus is maintained. Conversely, slowly adapting receptors fire continuously with the presence of the stimulus [79]. This also means that fast-adapting mechanoreceptors primarily capture dynamic signals, while slowly adapting receptors capture mainly static stimuli [44], [79].

Meissner’s corpuscles are fast-adapting mechanoreceptors that are highly sensitive to low frequency vibrations from 10-50Hz [80], [81]. Situated just beneath the epidermis, these tactile 11 receptors are mainly responsible for light touch. They are the most common form of mechanoreceptors of hairless (glabrous) skin, such as the finger pads or the lips, accounting for up to 40% of sensory innervation of a human hand [79]. Their rapidly adapting nature also allows the sensing of dynamic forms of touch (e.g., shape and textural changes) through the rate of skin deformation [82], [83]. Pacinian corpuscles are another type of rapidly adapting receptors. While present in fewer numbers compared to the other types of corpuscles (15% corpuscles in the hand), they respond to a wider range of high frequency signals from 40- 400 Hz propagating through tissues, and adapt more quickly than Meissner’s corpuscles. They are crucial in fine-texture discrimination or high-frequency (dynamic) vibration detection [79], [80], [84].

On the other hand, Merkel’s discs and Ruffini’s corpuscles are slowly adapting mechanoreceptors. Merkel’s discs accounts for 25% of the hand’s mechanoreceptors, and are particularly dense in areas such as the fingertips and lips [79]. Selective stimulation of these Merkel’s discs produces a sensation of light pressure and is important in the discrimination of edges, shapes and rough textures [79], [83]. The slowly adapting rate of these receptors makes them particularly well-suited for detecting static forms of pressure or low frequency dynamic skin deformations (<5Hz) [84]. Ruffini’s corpuscles, elongated, spindle-shaped structures, are typically oriented parallel to the skin’s stretch lines, making them particularly sensitive to skin stretch. They account for approximately 20% of mechanoreceptors found in a human hand. Although not well understood, it is believed that they respond to sustained pressure and allow for the perception of direction of object motion or force [79].

Since different types of mechanoreceptors have different rates of adaptations, thresholds of activation, locations, and depths in the skin, it is possible to activate different receptors based on their sensitivities to different features of electrical stimulus that passes through the skin. For instance, traditional high frequency vibration feedback commonly used in mobile devices are mainly detected by Pacinian corpuscles that are sensitive to dynamic, high frequency vibrations. This research in particular, has an interest in compression, but first and foremost, it is important to define the term ‘compression’ used in the context of this thesis.

2.2.2. Delineating the use of the term ‘Compression’

Mechanical forces applied to the body can take many forms. In the context of pressure-based haptic devices, one way of categorizing them are on the device actuators’ contact area; they could either 12 be localized, point-based (e.g., tapping, poking) or distributed (e.g., compressing, stroking). The current haptics literature uses the terms pressure, squeezing, and compression interchangeably to describe any encompassing, tangential force acting on a body surface, or constriction around a body part. However, we can imagine that such pressure sensations on a body part might involve not only tangential forces but might also involve shear forces when the device slides across the skin or when the skin stretches [35]. Gupta et al. attempts to delineate the term ‘squeezing’ to refer to ‘pressure sensations around a body part that consist of tangential and shear forces on it’ (described by their device that consists of a band tightening around a wrist), contrasting it to the term ‘compression’ that includes only normal forces (such as those with inflatable straps that pushes into the skin, e.g., a blood pressure cuff) [35].

While the mechanical distinction is important to note when discussing the biological involvement of this stimulus sensation—as described above, mechanoreceptors involved in both types of touch are different—depending on the intensity of each form of forces and the subsequent applications, their perception might not really be exclusive from each other [35]. The scope of this research does not include efforts to tease out the differences between terminology use and skin receptor response characteristics, but for clarity, in contrast to Gupta et al.’s definition, this research work will use the term ‘compression’ to describe the stimuli involving normal, tangential and shear forces acting on a body surface. The term ‘squeezing’ is not chosen here partly because it is commonly used to describe an action typically performed with one’s hands, and its colloquial use may be tied to emotional connotations, rather than a neutral mechanical stimulus. However, there could be instances where the term ‘squeeze’ may be adopted, for instance, when describing the stimulus characteristics to a participant, since it may be more easily comprehended (as it is commonly used in everyday language).

Tying it back to the discussion on sensory receptors, ‘compression’ under the definition of this work encompasses stimulation on the Merkel discs (which responds to sustained pressure), Ruffini endings (that are sensitive to skin stretch), and Meissner corpuscles (that respond to pressure changes at lower frequencies). The activation of these different classes of mechanoreceptors also play a role in the adaptation of the compression stimulus. As the compression begins on a body area, Meissner corpuscles that are rapidly adapting mechanoreceptors will immediately detect the dynamic changes in pressure on the skin. However, that signal will very quickly decay completely (within as little as 0.1 seconds), to the point where it may not be perceived in a binary way as the

13 compression is maintained [83], [84]. Conversely, slowly adapting receptors like the Ruffini endings and Merkel discs that are better at capturing static qualities of the pressure and shear stimuli associated with compression, will fire continuously given the presence of stimulus, albeit at a lower rate than when it first encounters the stimulus (hence could be detected as less perceptually intense after a period of time) [84]. The adaptation of these mechanoreceptors is the reason why we stop noticing the pressure from our clothing as long as it doesn’t change; our sensory system becomes accustomed to the signals. However, as soon as the stimulus changes, the mechanoreceptors will detect these minute changes and fire electrical impulses to the brain.

2.2.3. Psychophysics of Compression Perception

This review of the biology of the sense of touch would not be complete without a brief discussion of the psychophysics of touch perception, studied by varying the properties of a stimulus along certain physical dimensions systematically, and studying the effects on a subject’s experience [85]. It should be noted that the goal of this brief foray into the psychophysics literature is not to describe the detailed perceptual systems and processes of compression sensations, but rather to leverage these knowledge in the design of compression-based garment systems.

Touch psychophysical experiments are most commonly performed for vibrations or point pressures—rarely compression; likely confounded by the fact that the perception of compression is greatly affected by the method of delivery, involving a complex mix of sensations such as shear, normal force, tangential forces— with different devices providing different degrees of each of these forces, making the investigation particularly challenging. Compression differs from point pressures in that the sensations are more distributed and/or sustained. Therefore understanding the findings from point pressures, involving how mechanoreceptors are distributed over the body surface and how that affects pressure sensitivity may be important. Traditionally, this is done with the two- point discrimination (2PD) test, i.e., the ability to discern the minimal distance between two simultaneous stimuli as distinct [79], which allows the mapping of the sensitivity of various parts of the body to specific physical stimuli. For instance, the fingers and face are much more sensitive compared the limbs and torso [86], [87]. A comparison of 2PD thresholds for touch collected thorough the most heavily cited studies through the years are presented in Figure 2-1. Since the 2PD test is for individual point pressures, concrete claims cannot be translated directly to spatial discrimination of compression stimulus. However, we can draw loose parallels between the two when thinking about system design, since using larger point pressure actuator designs or higher 14 point pressure densities would eventually result in the convergence of both methods. More recently, Delazio et al. (2018) developed a body worn pneumatic jacket and determined how ‘accurate’ user’s perception of compression magnitude (delivered by the Force Jacket) was on different body locations [17]. They found that the shoulders were the most sensitive to pressure relative to other areas, while areas such as the back forearm, mid back, upper back, and upper chest were relatively insensitive. However, no absolute measures can be extracted from this study due to the free magnitude scaling method used to determine relative user perception of the stimuli.

Figure 2-1: Comparison of two-point discrimination for touch as measured by Weinstein (1968), Weber (1834), and Mancini et al. (2014) [88]–[90]. Figure obtained from Mancini et al. [90].

Given the differences in perception for different body locations, unique perceptual questions that can be further posed. Perhaps one of the most common is just-noticeable difference (JND), defined as the ‘smallest amount of difference a person needs to perceive two stimuli as different’[55]. From this end, Pohl et al. (2017), delivering pneumatic compression on the wrist, showed that unsurprisingly, the higher the difference in pressure, the more likely a stimulus pair was determined as unequal, and there was approximately a 2.77 offset to the base ratio for 95% of the participants to recognize the difference [34]. On the other hand, Gupta et al. (2017) used coiled SMAs to generate squeezing sensations on the wrist [35] and the JND resulted in a ratio of ΔLoad/ Load of 1.79, with more adherence to Weber’s Law compared to the 2.77 ratio in Pohl’s study. In a separate study by Mitsuda (2013), using a less conservative approach (50% estimate), also with a wrist-based compression device, the discriminative threshold pressures was found to range from ~2.25mmHg to ~3.83 mmHg [91]. Another common perceptual measure is the absolute detection 15 threshold, defined as the ‘smallest amount of intensity a person needs to detect a stimulus’ [55]. Pohl et al., using the same wrist-based compression system, showed that the subjects’ average detection threshold as 5.25 mmHg with a range of ~1.5 mmHg to ~11.93 mmHg. On the other hand, with Gupta et al.’s SMA wrist squeeze device, the mean threshold estimate was found to be 0.17kg with a mean response time of 1.4 seconds. However, the use of loads in kilograms to represent compression does not provide an intuitive nor translatable measure (the absolute pressure calculations are non-meaningful), and hence, also could not be directly compared with the results from Pohl’s study. These findings presented here however, are only used as broad guidelines in the development of compression systems in this dissertation work since the results obtained here are only focused on the wrist (in contrast to the goal of this study to involve more body locations), as well as utilized different compression devices/actuation strategies (likely resulting in different generated sensations than the soon-to-be constructed new compression garment platform).

To summarize, we observe an increased interest in psychophysical studies specifically geared towards compression in the recent years. However, the specific results can be difficult to put into perspective or applied directly to haptic device designs due to the different units of measure for absolute compression contact pressures, different method of compression delivery/actuation (which likely affects the perception of the stimuli itself), and the narrow focus of the perception only on the wrist/arm regions, and not on other areas of the body. The only certainty that can be extracted from the aforementioned studies is that the perception of compression varies on different body locations given the relative sensitivity/insensitivity of various body regions—hence the study of distributed on-body compression should involve body location as a variable to be investigated. Currently, with the exception of the work by Delazio et al. [17], all other compression-based psychophysical studies are limited to a single, localized body regions and there is a lack of information on how compression is perceived on different body locations. However, when studying user perception towards a compression stimulus, the characterization of psychophysical perception alone is not enough, as a haptic device has to be both functional and acceptable to users. Hence, this dissertation work considers subjective comfort/discomfort characteristics to also be important considerations for compression to function and be used as a form of interaction modality.

2.2.4. Discriminatory and Affective Touch Dimensions

In the previous sections 2.2.1-2.2.3, we briefly described the underlying cutaneous mechanisms for touch perception. However, in order for the signals to effectively be translated and processed by 16 the brain/nervous system, specific neural pathways have to be involved. We know that the sense of touch plays a role in exploration, identification, and manipulation tasks, but they also serve a separate function: touch is at the heart of many forms of social interactions and it can also be pleasant [1], [11], [92]. Recent evidence has shown that the neural organization of touch can take what Morrison et al. described as two functional, dissociable dimensions of touch. The first being the sensory-discriminative dimension and the second being the motivational-affective dimension [1], [92]. It is believed that the sensory-discriminative dimension encodes touch stimulus characteristics such as intensity and spatial localization, while the motivational-affective dimension encodes the valence (i.e. pleasantness and unpleasantness) and motivational relevance of the stimulus [92].

The nature of discriminative touch broadly involves fine touch (enables one to sense and localize a touch) and crude touch (allows one to sense a touch but it does not pinpoint the location where touch has occurred). These two types of touch involve different neural pathways. Fine touch, being a highly refined sensation developed in humans, is perceived and analyzed at the cortex level thorough the posterior column–medial lemniscus pathway (carried in the dorsal column) to the cerebral cortex of the brain [93]. On the other hand, crude touch is perceived below the cortical level, carried in the spinothalamic tract (through two adjacent pathways: lateral spinothalamic tract (transmits pain and temperature) or anterior spinothalamic tract (transmits firm pressure and crude touch) to the thalamus [1], [79]. It should be noted that this description is an oversimplification of the somatosensory system’s involvement and the actual neural pathways are much more complicated than the provided summary. However, the discriminative aspects of touch alone would not fully represent the sensation of touch in our everyday lives; one critical aspect is the affective dimension of touch—this is a much more complex perceptual experience compared to discriminative touch. Affective touch can be defined as the type of sensory touch information that elicits an emotional reaction and is most typically social in nature, commonly referred to as social touch (in contrast to other types of affective touch that could be sexual in nature). We will henceforth use the term social touch since it is the most relevant to this work’s application area.

There are two perspectives that social touch research can take; the first employs a bottom-up approach, starting from the vantage point of the haptic stimulus itself, working towards the sensory and neural processing of stimuli, and how the brain reacts to different touch stimuli [2], [3], [13], [94]. In somatosensory research, a common view is that touch is only mediated by fast-conducting

17 nerve fibers (myelinated Aβ afferent nerves) through the discriminative touch pathways discussed above [1]. However, research in the recent years have shown that the more interoceptive types of touch (i.e., signaling more ‘feeling’ rather than ‘sensing’ states) are mainly mediated by low- threshold unmyelinated C afferent fibers [1], [2]. Perhaps more crucially, the activation of these C- touch (CT) fibers was found to be linked with positive affect and pleasantness of touch stimuli [3], [13], [95]. However, these two (Aβ and CT) systems are likely intertwined—despite being partially dissociable —it is through integration that we can experience touch as a whole [92].

The second perspective takes a top-down approach, looking at the attribution and contextual factors that affect how social touch is defined. While the CT system is now seen as the major physiological platform for social touch, we know from experience and recent research that other ‘higher-order’ surrounding factors are also crucial aspects of social touch; this may include a wide range of contexts, the partner in the social touch exchange, stimulus types, delivery modalities, intent of the touch, and even age, gender, and cultural differences [11], [13]. For instance, Essick et al. (2010) found significant interactions between haptic ‘materials, sites, velocities, forces, and subject sex’ on pleasantness ratings, showing the complexities of the percept and raises questions as to whether it is even possible to produce an ‘affective map’ of body surface, similar to those of the discriminative dimensions in Figure 2-1 [3].

For more detailed review on the specific neural pathways associated with social touch, refer to [1], [11], [13], [92], [94]. In aggregate, this body of literature showcases the complexity of touch perception—it not only involves multiple somatosensory sub-modalities, but the boundaries of bottom-up (stimulus properties) vs. top-down influences (expectations and contextual cues) on touch have yet to be well defined. We know that the perceptual experience of touch itself is the product of converging input from various physical and cognitive processes; both from the physiological and psychological perspectives that invoke feelings and emotions, all while interacting with larger mediating factors in the environment including relationships, social context, and cultural factors to arrive at rich, affective information, emotions, and experiences.

Wearable Compression-Based Technology

In section 2.1, a broad survey on haptic technologies and their applications was presented, which showcased the wide range of possible capabilities of cutaneous tactile displays. Next, we summarize work related specifically to this research effort of compression-based wearable garment 18 systems. Increasingly, we see efforts in the robotics research community to move ‘earth-grounded’ haptic interfaces to smaller scale wearable interfaces, where ‘grounding/anchoring’ of the system (i.e., mechanism to counterbalance reaction forces) is closer to the point of stimulus application— hence making the devices more wearable/usable [45], [96]. While it is true in certain cases that by making devices more wearable, components of forces are progressively lost, the force decrements typically are kinesthetic in nature, and compression/pressure cutaneous displays are one of those that can still succeed without resorting to heavily grounded kinesthetic cues [47], [96]. This is especially true when the compression forces are presented circumferentially; coupled with the right design strategies, compressive forces can be effectively translated to targeted locations.

In particular, garment-based wearables are an ideal platform for delivering compression since they are (1) closely coupled to the skin for efficient stimulus transmission (usually necessary for physical cutaneous displays to be successful), (2) portable (by moving the ‘grounding/anchoring’ strategies closer to the stimulus application site), allowing a range of feasible form factors (e.g., 3D printed parts, textile-based solutions), (3) all while maximizing important wearability considerations e.g., discreetness and user comfort [41], [56]. Specifically, the garment form-factor provides access to wide areas of the body inconspicuously, maximizing the area of potential interaction while extending the reach of the system to any location on the body that is covered by the garment. This section details the unique qualities of compression stimuli and how those characteristics make them attractive for wearable haptic interfaces.

2.3.1. Information Display/Notifications

In current commercial devices, cutaneous tactile stimulation is most likely encountered as a form of information display. As such, many research investigations involving compression are also as a form of information communication modality. In contrast to vibration feedback that we typically encounter, compression is said to be less attention-demanding (making it suitable for background feedback), more pleasant than vibrotactile approaches [18], [35], [97], and can even slowly ramp up/down to bring attention to the user as an indication of a change in status [34]. These signals can take a wide range of attention capture, from subtle, to inhibiting, and even forceful (to indicate the urgency/status of the information), through various compression patterns [97]. Pohl et al. (2017) showed this through their Squeezeback wrist device with psychophysical experiments, comparing it also to vibrotactile strategies. They found that compression and vibration feedback performed similarly as a notification mechanism, the only exception being that of reaction times— 19 compression feedback took slightly longer for users to react to, compared to vibration feedback [97]. Gupta et al. (2017) performed a similar study but using different actuation mechanism (SMAs), and found that their wrist device could generate up to 4 distinct compression levels, works well in distracted use-cases, and was successful in communicating progress of an activity [35]. Chernyshov et al. (2018) also demonstrated the use of compression in a ring device to present various information types, including a progress bar, timing reminder system, and information on a user’s bank account balance [24]. Compression stimuli is also similar to attention-capture behaviors performed by humans—grabbing someone’s wrists to get someone’s attention. Baumann et al. (2010) demonstrated this with their ServoSqueeze wrist device, demonstrating how participants were able to use compression to convey a range of socially-gradable attention capture expressions [98].

From these studies, we see that compression feedback has potential in being used as an information display. However, because of the nature of the stimulus type, as well as how pressure stimuli are detected by our mechanoreceptors (slower reaction times), they may be less suitable for urgent alerts requiring immediate response—but rather, as a form of ambient display, with information presented in the periphery of human attention. Another thing to note as well is that all of these compression-based notification devices are centered on the hand/wrist/forearm regions in a strap- or band-like form factor, likely due to the familiarity of those body locations afford for wearable devices (e.g. similar to a watch). There is currently a lack of focus on investigating compressive forces on larger, distributed body regions; this research that concentrates on body-scale garments that provide access to multiple body areas could maximize the area of potential interaction or the amount of haptic information presented.

2.3.2. Affect Communication and Social Presence

Previously we discussed the importance of social touch and how it can profoundly affect social interactions within a community. Different forms of touches may bring different meanings, as demonstrated by an extensive diary study done by Jones and Yarborough, who showed 12 distinct categories of meaning that different types of touches can bring [99]. More recently in the realm of human-computer interaction, many researchers are becoming increasingly interested in the idea of mediated social touch—where people engage in social touch with each other over a distance by means of haptic technology [5], [10]. There are generally two goals of this type of research: either to find similarities between human-human social touch and mediated social touch and reproduce 20 those haptic stimulation effects, and/or to see how haptic stimulation affects mediated interpersonal communication. A large portion of these studies focus on the communication of affect—either positive/negative affective states or in the form of discrete emotions. These were perhaps motivated by Hertenstein et al.’s work [68] and later others [28], [29] that showed discrete emotions can be communicated through touch alone based on varying tactile behaviors. Later, understanding the potential link between the role of touch and emotional communication, researchers started to investigate if haptic technologies can be used to mediate the communication of affect and emotions. Smith and MacLean (2006), through a purely haptic link using virtual models rendered on simple knobs [31] and Huisman (2013) through an instrumented vibrotactile array sleeve [100], showed that discrete emotions can indeed be communicated through haptic devices alone. However, the effectiveness of the communication depends largely on the type of haptic feedback used, body locations stimulated, and relationship between communication partners.

Where compression as a stimulus really shines in the realm of mediated social touch is its unique qualities of being very similar to common human behaviors such as a hug, a handshake, or a grab of the wrist to get someone’s attention. With that, many mediated social touch devices have been replicated to provide pleasant sensations and communicate positive intent such as affection and love. It is also commonly used to enrich or support online communication and communication between users separated by distance. Some examples include HaptiHug that represents a hug in a virtual world [101], Huggy Pajama to deliver mediated hug during remote communication between parent and a child [74], [102], and Hug Over a Distance for expressing intimacy between long- distance couples [72]. The key driving force in this area of study is the ideas that touch carries huge emotional significance [25] and the eventual demonstration that interpersonal touch does not necessarily have to be direct to constitute as being social [13]. One thing to note is that in contrast to the use of compression for notifications presented in section 2.3.1, these affect-related devices takes much more diverse form factors, extending away from the wrist to other regions of the body in softer, wearable form factors, presumably to facilitate user acceptance and comfort, replicating common human behaviors like a hug, while providing an intimate form of feedback.

Beyond the communication of affect, compression-based mediated social touch technologies may also be used to enhance feelings of social presence and task performance. Yarosh et al. (2017) implemented a Squeezeband device to investigate the supplementation of haptic communication in a video chat scenario using a ShareTable system, while performing a task in either a high emotion

21 or low emotion scenario. They found that the haptic compression gesture support may be particularly effective in high emotion tasks perhaps through easing mental and physical demands [75]. Another work related to social presence was done by Wang and Quek (2012), utilizing compression stimulation in a storytelling scenario. Participants who attributed the compression to the storyteller during emotional moments in the story felt closer to the storyteller than those who attributed the compression feedback to the storyteller’s autonomic arousal [60]. In a separate study by the same authors using a different form of squeeze device, but also in a storytelling scenario, they showed the compression stimulation was successful in reinforcing the meaning of the situation, by reducing general negative mood (especially sadness) and to reinforce feelings of joy [19].

Although there is huge potential in using compression in various mediated social touch applications, it should be noted that there are differences between actual human touch and mediated touch technologies [5], [10]. This includes the fact that mediated touch is not necessary reciprocal and can occur asynchronously. For instance, a user can ‘pre-record’ certain touch protocols and send them to a receiver at a pre-determined time in the future. Further, in many applications where mediated touch is used, the touch sensations are delivered over a distance (where sender and receiver are not co-located). This in turn causes mediated touch to perhaps be less sensorially rich than actual social touch; this includes not only the physical cutaneous sensations, but contextual cues and other modalities (audio/visual) which may or may not be present in mediated social touch situations. Therefore, all of these factors have to be taken into consideration when designing and implementing mediated social touch technologies to communicate or modulate affect.

2.3.3. Media Augmentation and Entertainment

Another area of focus that compression has received some attention is the use of compression stimulation in media augmentation. This contains some overlap with the previous section in using compression for interpersonal communication, especially when coupled with another system such as video call technologies, interaction with characters in virtual environments [5], [103], and even enhancing/inducing emotions in multimodal applications such as movie viewing [27], [40], gaming in virtual environments [17], musical experiences [104], and storytelling [19]. However, a majority of these research efforts are skewed towards vibrotactile approaches and rarely on compression, even though compression is arguably a more realistic proxy for human touch.

Currently, there are some empirical studies regarding the use of compression as a haptic modality 22 for entertainment purposes or to increase immersion in virtual environments; a portion of these studies are focused on developing ‘feel effects’ library. Feel effects are generated haptic patterns that are driven by a specific haptic technology through an explicit pairing between a meaningful event or linguistic phrase, first introduced by Israr at al. [105]. The first research is the work by Delazio et al. (2018) known as the Force Jacket, a garment that delivers pneumatically actuated compression stimulus to the upper body. They showed the possibility of the jacket to produce a library of 14 feel effects including hugs, muscle enhancement, snake slithering around the body, snowball hits et cetera, and demonstrated potential applications of their device in VR applications [17]. The second is PneuSleeve, by Zhu et al. (2020) that demonstrated the expressivity of the developed forearm sleeve (capable of compression, skin stretch, and vibration), with 23 feel effects for a variety of purposes such as those simulating acceleration, directional rotation, navigation, and other special effects such as snake crawling and breathing [38]. These studies demonstrate the expressiveness of compression technologies, especially on certain effects that cannot be replicated through vibration alone, such as hugs, breathing, and ‘waves’. The next step in this research area will require multi-sensory integration and how these produced feel effects can further augment visuo-audio media and enhance user experiences.

2.3.4. Persuasive Touch in Health

Previously, we discussed the use of haptics in communication for information transfer, emotional communication, media augmentation, as well as social mediated touch. Sitting within the intersection of these fields lie another form of approach—using haptics for attitudinal or behavioral change [10]. This approach largely intersects with persuasive computing, a field that broadly describes technology that is aimed at changing the attitudes and/or behaviors of users through persuasion and social influence, without the use of coercion or deception [106]. The role of social touch in persuasion has been explored in many different ways including understanding social influences, consumer behavior, and promoting health or safety changes [10]. Many have also taken this one step further, creating technologies that are persuasive, aiming to forge new habits, or change the attitudes of users. For example, digital health coaching or using haptic and tactile information to improve driving safety [107].

Perhaps one of the first studies on direct social touch that received much attention was by Crusco and Wetzel (1984) who showed that a relatively unobtrusive touch on the hand or shoulder resulted in a significant increase in tipping by patrons, with the researchers dubbing this as the ‘Midas Touch’ 23

[12]. A brief touch to the hand, arm, or shoulder can positively affect attitudes towards the sender [108], [109], or result in behavior changes including willingness to participate in surveys [110] and even obtaining a free bus ride [111]. This social touch effect in eliciting changes in attitudes/ behaviors also extends to healthcare; touch has been found to be effective in comforting those who are ill, reducing their perception of pain, and positively influencing mood [112]–[114]. Touch has also been used in various forms of therapeutic interventions including massage therapy and occupational therapy. In this context, the use of compression is known as deep touch pressure (DTP), involving the delivery of surface pressure on the body that induces mechanical deformations of skin receptors and underlying connective tissue structures (such as the periosteum and fascia), affecting deep receptors in the skin [115]. This form of pressure can be exerted in circumstances such as firm touching, holding, swaddling, hugging, and squeezing—either delivered by a person or a system—which are all related, in some form, to compression.

In the context of occupational therapy, DTP is commonly used as an intervention with clinical populations experiencing sensory processing disorder, attention deficit hyperactive disorder, learning disabilities, and autism spectrum disorder. It could involve a variety of system types, from large contraptions such as the Grandin Hug Machine, to wearable systems such as weighted vests and pneumatic-based compression garments [21], [70], [115], [116]. DTP is said to provide benefits physically, mentally, and emotionally, by encouraging stress reduction, relief of muscle tension, relaxation in the state of alertness, reduction of mental anxiety, and an overall improved feeling of well-being [9]. The administration of DTP is grounded in research based on Sensory Integration (SI) theory whereby calming effects are believed to be a consequence of modulation of how the central nervous system processes sensory information [117], [118]. All these taken together shows the potential for using haptics—in this case, compression stimulation—to not only modulate affect, but potentially induce attitude or behavioral change.

Currently, most persuasive computing systems convey information explicitly to the user and require users to consciously decide to act upon the information (e.g. fitness trackers, mobile applications for meditation). This is a form of persuasive technology, but the persuasion is done indirectly since it still requires conscious effort on the user’s part for a change to occur. Conversely, a more direct form of persuasion would be to induce changes without conscious action from the user. As mentioned, compression has been demonstrated as an effective therapeutic intervention, demonstrating the potential for a direct link between stimulation (e.g. compression) and

24 attitude/behavioral changes (e.g. calming effect) [11]. If coupled with a feedback mechanism that can autonomously detect context or the onset of triggers, such actuatable compression systems have huge potential for growth as a novel form of persuasive computing. While the demonstration of this idea specifically is not within the scope of this dissertation, it remains an important motivation of this research; a long-term direction of this work could enable the extension of the use of haptics beyond purely affective purposes and into the realm of direct persuasion.

In summary, to truly maximize the use of compression as an interaction modality, be it for communication or modulating affect, the effects of compression on the human body must be better understood. Broadly, the experience of compression can be modulated with varying inputs including application location, intensity, rate, and duration, through the use of different actuation mechanisms and technological approaches. Therefore, this work aims to understand the various factors influencing compression stimulation on the body for a desired experience through the use of wearable systems.

Compression Technology Form Factors & Actuation Mechanisms

From a technology standpoint, current work in compression-based wearables take two trajectories. The first includes devices that present compression stimuli on smaller body areas, most commonly the hands/wrists/forearm [24], [34], [35], [38], [75], [119]. Such localized haptic displays have advantages including user/social acceptability (since it is on a body location that people are familiar with placing devices), as well as technology simplicity (since it is likely to require less operating power). However, they are in turn, more limited in their display/interactive/persuasive abilities, since their inherent localized nature typically only provide one point of contact, reducing the amount and type of information that can be communicated. Therefore, this work aims to investigate larger body areas—towards the goal of body-scale devices—by using a garment form factor to deliver distributed compression stimuli. Garments are ideal for creating compression since it provides a large direct-acting surface area and has close proximity to the body (which holds the potential for a far greater amount and type of information to be communicated [56]), while also allowing the balance of key wearability considerations.

The second group of compression-based devices follow this train of thought by presenting compression on larger body areas, typically in the form of jackets/vests [17], [63], [72], [102]. Current garment-based compression systems are typically either (1) passive, non-adjustable 25 garments exerting compression through added weights or stretch (e.g. weighted vests, elastic clothing), or (2) dynamic garments constricting the user, as summarized by the technologies presented in Table 2-1 [120], [121]. Passive, non-adjustable garments offer relatively good mobility and portability, but does not afford active/dynamic actuation and therefore will not be considered. Active garments offering dynamic actuation can be largely categorized into three groups based on their mechanisms of action. Traditional methods inspired by robotic systems typically include either (1) servo motors through tensioning of straps or moving plates, or (2) inflatable pockets/ bladders driven by pneumatic systems. However, servo-driven solutions are typically bulky and lack wearability considerations since they require positioning of hardware (e.g., motors) for successful tensioning of straps and are thus usually only considered for very localized regions such as the wrist and arms (Table 2-1a). On the other hand, pneumatic systems are much more flexible in terms of implementation and the creation of pneumatic bladders are relatively established, therefore is usually the selected choice for targeting both localized and distributed body regions, from the wrists and arms to torso and shoulder regions (Table 2-1b). However, with the exception of a few solutions [74], [122], pneumatically driven systems are generally encumbered by the noise or portability of air pumps needed to drive system inflation. Therefore, a potential solution to close this research gap is to develop computer-mediated compression garments that are dynamic, low- profile, and remotely controllable.

A newer class of materials, known as shape memory alloys (SMAs), are a particularly appealing solution as a method for providing compression in the recent years. They are a class of active materials (i.e., soft robotic artificial muscles) that feature the ability to repeatedly change shape when actuated [123]. When designed to be in a coil configuration, their large active displacements can be used as a mechanism of donning and doffing loose garments while retaining the ability to dynamically apply compression when actuated. Perhaps more importantly, they can be easily and unobtrusively integrated into typical textile structures, and actuated using an electrical current, making them an attractive actuation mechanism for applying compression stimuli. However, many of these SMA technologies now are only focused on applying localized compression on smaller areas of the body such as the wrists or fingers (Table 2-1c). The only exception being the work done by Duvall et al. (2016) with a child-sized garment concept proposed to apply compression on the torso, but the technology has yet to be tested on humans.

Summary Table of Wearable Compression Technologies

The following tables (Table 2-1(a)- (c)) provide a list of wearable compression-based interfaces, subdivided based on their technology’s actuation mechanism. The information presented in the tables provide an abbreviated view of each research studies’ target body locations, type of haptic stimuli, haptic actuators and their power source, as well as the intended or targeted application area as identified by the authors. Note that there are other forms of actuation technologies not covered in these tables (due to the lack of widespread adoption or technology maturity), for instance, other soft robotic actuation technologies such as fluidic fabric muscle sheets introduced earlier as the work by Zhu et al. [38].

(a) Servo motor solutions

Servo Motors Purpose or Research Form Body Haptic Haptic Power Year Potential Work Factor Locations Stimuli Actuators Source Applications Cord- Interpersonal tensioning communication, Pezent et al. 2019 Bracelet Wrist Compression motor, Electric Information (Tasbi) [124] vibrotactile display, actuators Entertainment Moving Chinello et al. Moving Mimic squeeze plates driven 2014 (HapBand) plates Forearm Force Electric forces by by servo [125] (hardware) separate glove motors Servo motor Recreating Stanley & Watch- Force, Twist driven sensations 2011 Kuchenbecker type Wrist (skin stretch), linkages, Electric experienced in [119] wristband Drag, Tap tensioning human straps interactions Baumann et al. Watch- Servo with Attention- 2010 (ServoSqueeze) type Wrist Compression tensioning Electric getting [98] wristband straps expressions Tensioning straps, Chest, Force and Emotional Tsetserukou et Body- Speaker, Back, Compression, augmentation 2009 al. (iFeel_IM!) worn Vibrotactile Electric Abdomen, Vibration, during online [126] device actuators, Sides Temperature communication Peltier elements

(b) Pneumatically driven inflatables

Inflatables Purpose or Research Form Body Haptic Haptic Power Year Potential Work Factor Locations Stimuli Actuators Source Applications Zhu et al. Compression, Expressiveness Fluidic fabric 2020 (PneuSleeve) Sleeve Forearm Skin Stretch, Pneumatic in simulating muscle sheets [38] Vibration feel effects Force (individually Young et al. Thermoplastic Information addressable 2019 (BellowBand) Bracelet Wrist polyurethane Pneumatic display, areas) and [127] bellows Entertainment Compression, Vibration Shoulder, Expressiveness Forearm, Force in simulating Delazio et al. Upper arm, (individually Polyurethane feel effects, 2018 (ForceJacket) Jacket Torso Pneumatic addressable bladders Media [17] (front, areas) augmentation, back, Entertainment sides) Pohl et al. Modified Information 2017 (Squeezeback) Bracelet Wrist Compression blood Pneumatic display, [34] pressure cuff notifications Raitor et al. Thermoplastic Information 2017 (WRAP) Bracelet Wrist Compression polyurethane Pneumatic display, [128] bladder Guidance Clinical Reynolds et Intervention Pneumatic al. (VayuVest- (promote 2015 Jacket Torso Compression Air pouches (Hand Commercial calmness)- Pump) Device) [129] Deep Pressure Therapy Force Information He et al. (individually Silicone-cast display, 2015 (PneuHaptic) Band Forearm addressable pneumatic Pneumatic Interpersonal [130] areas) and actuators communication Compression Modified Force display 2013 Mitsuda [91] Bracelet Forearm Compression blood Pneumatic for robotic pressure cuff applications Force Teh et al. (individually Interpersonal (Huggy Chest, addressable 2012 Jacket Air pouches Pneumatic (remote) Pajama) [74], Shoulders areas) and communication [102] Compression, Warmth Clinical Intervention T-Jacket Torso and (promote 2011 (Commercial Jacket Compression Air pouches Pneumatic Shoulders calmness)- Device) [122] Deep Pressure Therapy

SMA Systems Purpose or Research Form Body Haptic Haptic Power Year Potential Work Factor Locations Stimuli Actuators Source Applications Forearm, Long- Force through Information Kentaro et al. Upper 2019 sleeve T- clothing SMA coils Electric display, [131] arm, shirt deformation notifications Shoulders Variable Stretching, Hamdan et al. (Head, Pressing, Information Skin 2019 (Springlets) Wrist, Pulling, SMA coils Electric display, interface [132] Hands, Dragging, Entertainment Fingers) Expanding Affect Papadopoulou Compression modulation 2019 et al. (Affective Sleeve Forearm SMA wires Electric and warmth (promote Sleeve) [30] calmness) Rings Finger Information Chernyshov et 2018 placed on (each Compression SMA wires Electric display, al. [24] a glove phalange) notifications Gupta et al. Information 2017 (HapticClench) Bracelet Wrist Compression SMA coils Electric display, [35] notifications Yarosh et al. Collaborative Compression 2017 (SqueezeBands) Band Hand SMA coils Electric and social and warmth [75] computing Clinical Intervention Duvall et al. (promote 2016 (Hugging Vest) Jacket Torso Compression SMA coils Electric calmness)- Deep [133] Pressure Therapy Wang & Quek Augmenting and Upper 2012 (Touch & Talk) Armband Compression SMA wires Electric reinforcing Arm [19] experiences SMA with Band Variable non-elastic (modular), (Head, Compression, straps, Suhonen et al. Interpersonal 2012 Watch- Wrist, Vibrotactile, vibrotactile Electric [18] communication type Hands, Thermal actuators, wristband Fingers) Peltier elements

Research Gaps in the Study of Compression-based Wearables

As shown in previous sections, studies using compression as a haptic modality and its affective abilities intersect many application areas including human-computer interaction, information communication, affective haptics, wearable technology, and healthcare interventions. Leveraging the broad body of previous work in this area, this dissertation is driven by several identified gaps

29 in the current research landscape; each of these gaps can be mapped to the major contributions of this dissertation (as presented in Section 1.2). The identified research gaps and their associated research objectives are as follows.

(1) There is a lack of focus on distributing compressive forces on larger body areas in a controllable and unobtrusive manner, presented through non-bulky (e.g., non-servo/pneumatic) garment form factors

A large majority of past research on the use of compression as a haptic interaction modality are largely limited to localized body regions, targeting a relatively small surface area or only single body location (e.g., arm/wrist-based technologies) [34], [119]. The involvement of larger and/or more body locations in the presentation of compression stimuli is interesting to consider since more information can be transmitted given the additional degrees of freedom in the haptic display (i.e., body location as an added parameter that can be used to encode information, other than intensity/ duration/pattern) [56]. Garment-based systems are an ideal platform for creating compression on the body since they provide access to wide areas of the body in a direct yet discreet manner, and in turn, maximizing the area of potential interaction to involve any body location that is covered by the garment. However, current research on garment-based compression systems are limited to traditional actuation schemes (e.g., servos, hydraulics, pneumatics), that generally involve hard, electronic or actuation components.

Specifically, a key idea in this dissertation is to create an innovative garment platform that abandons traditional wearable actuation schemes and instead shifts toward the use of embedded active materials (i.e., soft robotic artificial muscles) that feature the ability to repeatedly change shape when actuated. Shape memory alloy (SMA) actuators have been identified as an appealing actuation solution but most SMA-based strategies have only focused on wrist-based devices. Hence, this work is motivated by the research gaps observed in terms of technology development, whereby there is a lack of focus on using garment form factors to distribute compressive forces onto larger body areas (i.e., upper body regions such as torso, shoulders, arms), through SMA- integrated garment technologies that can deliver compression inputs in a controllable manner.

However, it should be noted that there is an inherent challenge in presenting compression on other body areas (e.g., torso and shoulders) since anatomically, they are less simple than the wrist/arm. The forearm circumference and tissue distribution are relatively uniform, hence the limb can be

30 simplified and modeled as a cylindrical object. In contrast, there are significant variability in body tissue distribution in the torso and applying compression on the shoulders require effective ‘grounding/anchoring’ techniques to ensure the compression is effectively translated to the target body locations. Further, the complex geometry and tissue properties of the shoulder regions (combination of bony areas and soft tissues) makes applying circumferential compression on that location particularly challenging as compared to the wrist/arm regions. While there are possible strategies to counteract these concerns, the design challenges are non-trivial.

Research Objective 1: Investigation and development of compression garment technology that are capable of delivering controllable dynamic compression inputs (location, intensity, duration, pattern) on distributed, upper body areas, delivered through soft garment form factors (i.e., soft robotic actuation technology), that ultimately enables new types of on-body haptic experiences.

The soft form factor allows a reduction of mechanical impedance between the technology and a user, which in turn enhances user comfort and system wearability. This will be the first investigation towards using the selected soft robotic actuation mechanisms coupled with soft garment structures to deliver compression sensations on large body areas inconspicuously, and the technology can further serve as a research tool that enables the experiential and functional evaluation of on-body compression.

(2) There is a need to better understand the subjective, experiential effects of compression stimulation given varying compression parameters on various upper body locations

While there is considerable evidence showing the benefits of strategically applied compression on the body, both as communication displays and affect modulation tools, there is still a lack of rigorous studies to understand the subjective effects of compression stimulation especially on larger body areas like the torso. The experience of compression can be modulated with varying inputs including compression application location, intensity, rate, and duration, but how these parameters impact user perception and preferences are yet to be defined. Therefore, to truly maximize the use of compression in various application spaces, this dissertation work also seeks to investigate the relationship between delivered compression parameters and experienced perceptual effects/ subjective preferences, possibly determining if there are proper compression stimulation

31 parameters for a desired experience. We are also interested in understanding user variability, including if/whether compression is universally enjoyed or potentially bifurcated into affinity and aversion groups.

Research Objective 2: Assess and identify the subjective, experiential effects of computer- mediated compression (including user perception and preferences), given varying compression stimuli parameters presented on various upper body locations as afforded by the designed soft- robotic compression actuated garment technologies.

(3) There is a need to map the design space for SMA-based, compression-mediated emotion communication

We previously discussed the wide array of application spaces that compression-based technologies can take, one of which is their use in emotional communication. Currently, most wearable, upper body haptic-mediated social touch technologies for affect communication are largely limited to communicating general positive affect [72], [74], [102], [39], ignoring a wide range of other possible emotions that may be of interest/use despite considerable evidence showing that distinct emotions (e.g., fear, anger, happiness, love) can be communicated via direct (human-human) touch alone [68], [134]. Further, when haptic-mediated emotional communication is studied, how exactly the emotions are communicated through mediated touch are rarely clarified (i.e., it is not only important to understand the haptic strategies people use to communicate emotions through haptic devices, but also the mental models associated with the selected strategies). A given understanding on both how and why users choose to communicate various emotions through a mediated affective garment, capable of presenting compressive forces of varying inputs (through SMA-driven technology) is important and currently not well understood. Hence, another facet of this dissertation work aims to map the design space for SMA-based compression-mediated in affect communication by explicitly involving users in gathering compression haptic strategies for emotional communication, while attempting to parse thought processes and mental models used during the process, to develop design strategies for future compression-based haptic communication applications.

Research Objective 3: Characterize user expectations and mental models associated with computer-mediated compression as a communication and display modality and develop design strategies for future haptic communication applications that utilize the designed soft robotic compression garment technologies.

(4) The use of SMA-actuated, compression-based interactive systems to modulate emotions is still in its relative infancy, hence there is a need to demonstrate the potential and/or effectiveness of these types of technology as an affective haptic system

In the previous sections, we discussed past work that have shown the ability of users to use mediated social touch technologies to communicate emotions, signal positive affective states and intent, as well as improving social presence in certain use scenarios. Further, various healthcare applications have also shown the benefits of strategically-applied compression by decreasing anxiety, promoting feelings of calmness, improving alertness, and positively influencing the overall well-being of users [9], [20]–[22]. Knowing the potential of using compression-based interactive systems to augment/enhance/modulate affect in a variety of applications, this work also seeks to investigate the use of compression-based wearables from an affective-based application (functional) standpoint and demonstrate the potential for use of such technologies in future use cases or research avenues.

Research Objective 4: Demonstrate the potential of using the designed soft robotic compression garment technologies in affect modulation applications through a compression- guided meditation task.

Chapter 3

3. User-Centered, Iterative Design of Soft Robotic Actuated Compression Garment Technology

Compression-based wearable systems with varying modes of actuation and applications were identified in Chapter 2. Although they are used in many different fields, the grounded design principles for upper body compression garments have yet to be clarified. This chapter involves translating technologies and approaches from other soft robotic researchers to innovative soft robotic garment applications that enable novel haptic experiences, while synthesizing best practices and insights for the design of upper body compression garments. Specifically, we present the engineering and design of dynamically controllable wearable compression systems capable of delivering compression sensations in a controllable manner across large areas of the body, while balancing elements of wearability and usability through the use of soft garment form factors. The developed technology also is designed to act as a research platform, enabling the experiential and functional application/evaluation of on-body compression.

A human-centered design approach, involving iterative design and user evaluation throughout the development process was employed so that a well-rounded perspective of the design space can be captured—balancing between user desires/needs and technology viability/feasibility [135]. Since this is a wearable system that deals with technology-integrated components—towards the aim of being worn unobtrusively as clothing—there are plenty of considerations in regard to how the designed system could affect users’ experience with the world and their own bodies [136]. However, designing wearable haptic systems that fit well on people’s bodies comfortably while maintaining technology functionality is a non-trivial task as our bodies are soft, geometrically complex, and in constant motion [41], [56], [136]. Hence, a key idea is to involve users directly in the design process to understand the relationships between the human body and the wearable technology, as well as the impact of this active relationship on a user’s experience.

In this chapter, we present an end-to-end design and development process of soft robotic actuated compression garment technology 1 through several user studies, starting with a broad approach of first understanding how users experience compression on various areas on the upper body with a simple, (low fidelity) adjustable passive compression garment, followed by a technology-integrated (medium fidelity) compression garment, and finally a refined (high fidelity) remotely-controllable dynamic compression garment that could be used as a research tool in understanding the effects of computer-mediated compression on the body. Since there is also currently a knowledge gap in terms of the ways soft robotic technology can be integrated into garment structures to effectively leverage their actuation properties as a compression interaction modality, this chapter documents both the design process and the synthesized the engineering/design innovations, strategies, and best practices related to the coupling of soft robotic actuation technologies with a soft goods platform.

1 The iterative prototyping effort of the SMA-based compression garments would not have been possible without the support from other members of the Wearable Technology Lab, namely J. Walter Lee (M.S. graduate of UMN Human Factors and Ergonomics program), Simon Ozbek (M.S. graduate of UMN Human Factors and Ergonomics program), and Crystal Compton (Ph.D. candidate of UMN Apparel Studies program). They played a crucial role in helping pattern and sew the medium- and high-fidelity prototypes, manufacturing SMA actuators, providing note-taking support during the user studies, as well as copy-editing support for subsequent conference publications.

All garment iterations demonstrated in the subsequent studies utilize fabric structures, coupled with actuation technologies, to deliver compression sensations to a user. This ‘soft’ device structure is a unique aspect of all the systems developed in this dissertation; a soft interface that features a smaller mechanical impedance differential between the device and a user should provide better fit, better transmission of haptic sensations, improved wearability (e.g., far less bulk/stiffness than a comparable pneumatic system), and ultimately more user-friendly for eventual deployment beyond laboratory experiments—all of which are crucial facets of user perceptions and experiences towards a designed haptic wearable system [38], [41].

Exploratory Study with Low Fidelity Prototype

The goal in this first exploratory study is to gather rapid feedback from users with the aim of understanding the design space, and to establish design requirements for dynamic compression garments. In particular, we are interested in investigating control mechanisms capable of providing remotely controllable dynamic compression inputs (e.g., location, intensity, duration, pattern), while maintaining wearability elements crucial to user experience (e.g., unobtrusiveness, comfort, portability, aesthetics). Instead of immediately investing in complex actuation technologies, this exploratory study first utilizes an upper body garment, capable of providing variable compression sensations through simple manually-adjustable Velcro straps and constructed with readily- available materials—henceforth referred to as low fidelity prototype, as a first step in understanding users’ subjective experiences with applied compression, while gathering information on areas that require particular attention when it comes to compression garment design [137].

3.1.1. Low Fidelity Test Garment Design

The low-fidelity garment’s major design criterion is the flexibility of compression application— with an adjustable garment, various compression locations and intensities can be applied. The adjustable test garment (Figure 3-1) was designed to allow maximal adjustability in the application of passively applied compression on two different body regions: the torso and shoulders. These locations were selected because they make up most of the upper body surface area of the torso and most of the current compression garment designs also target these locations [17], [122].

Figure 3-1: Low-fidelity garment- Front view (Left), Side view (Middle), Back View (Right). Hook-and- loop straps are anchored strategically on varying garment locations to allow for maximal adjustability and flexible application of compression on various body regions

This low fidelity garment uses Hook and loop (Velcro®) straps strategically anchored throughout the garment and were free to move in various orientations—it is through the tightening of the straps that induces tensioning of the fabric, that compression is applied the body. The garment base was constructed using a non-stretch woven canvas fabric to ensure that: (1) the forces induced by tightening of the strap were directly translated into tensioning of the non-stretch fabric (i.e., if a stretchy fabric was used, the tightening of the Velcro straps will pull the stretchy fabric and compression forces will not be efficiently translated until the stretchy fabric reaches its stretch limits and induce a blocking force), and (2) that the compression intensities are relatively constant across participants given the stability and structure of the fabric and garment. Since the garment was designed to be adjustable, the size of the garment was constructed to approximate an average adult male participant (Size S-M)—the upper limit of participants we were intending to recruit.

With the hook and loop straps anchored throughout the garment—that when tightened, induce fabric tension, compressive forces can be applied in various orientations on the body. For compressing the trunk of body, straps were oriented perpendicular to the garment’s side seams to provide circumferential compression on the torso. For the shoulder areas, straps were anchored on the shoulder seams (top of shoulders) and extended vertically, running along either front of the chest or shoulder blades. The strap orientation on the front and back shoulder regions can afford many compression vectors, here 4 distinct compression vectors were identified: (a) ‘straight’, (b) ‘one-side crossed’, (c) ‘diagonal crossed’, and (d) ‘mixed’ (Figure 3-2), as they all provide varying compression profiles. In the ‘straight’ shoulder condition, the straps were oriented vertically (parallel to the arm), similar to a backpack strap. In the ‘one-side crossed’, the two straps on each side of the body crossed each other without crossing over the body’s midline. In ‘diagonal crossed’, 37 straps on each side of the body crossed the body’s midline diagonally. In the ‘mixed’ condition, the outer strap (closest to arm scye) was vertical, and the inner strap was crossed. Markers were placed various locations on the garment to ensure consistency of strap angles across participants.

Figure 3-2: Shoulder test conditions with changing compression vectors: (a) ‘Straight’, (b) ‘One-side crossed’, (c) ‘Diagonal crossed’, and (d) ‘Mixed’

One big challenge in this study was the method for applying uniform compression across subjects of various body shapes and sizes, as well as accommodating contoured body surfaces. To overcome this, after donning the passive garment, the straps were adjusted to provide a baseline loose fit (i.e., conforms to the body without exerting pressure points). Any pressure points were self-identified by participants and corrected by loosening corresponding straps; the straps were graded with 0.5- inch increments. In each test condition, the straps were subsequently tightened by 1.0 inch on both sides from the baseline. This ensures that all participants were given approximately same amounts of compression in each situation.

3.1.2. Low Fidelity Prototype User Study Methodology

A within-subjects design was used in this pilot study with 5 participants (2 males, 3 females; age range= 22-30; average age= 25.4 years old). Each participant was given a cotton jersey long sleeved T-shirt to wear during the study to ensure similar baseline compression. Each participant was exposed to 6 test conditions: ‘torso’, ‘straight’, ‘one-side crossed’, ‘two-sides crossed’, ‘mixed’, and an informant design. The study was pseudo-randomized; the torso compression condition had to be performed first because the shoulder test conditions were contingent upon anchoring of the trunk of body (i.e., without fitting the garment on the trunk, the garment will ride up as the shoulder straps were adjusted). The next four trials were the four aforementioned shoulder conditions (Figure 3-2) in a randomized order. During each of the shoulder test conditions, both front and back straps were adjusted to the same orientation. Note that during the shoulder test conditions, the torso remained compressed. The final condition was an informant design exercise whereby the participants sketched out what they wanted or instructed the moderator to move the straps (both 38 orientation and intensity) until it gave the most comfortable and satisfactory compression sensation. This was also an attempt to capture the variability in subjective perceptions in comfort between participants to better inform future designs.

Each of these six test conditions lasted for 2 minutes (this time frame was selected since participants would have had enough time to feel changes in perception of sensations and had a chance to accommodate to the new sensations and form an opinion on their preference), during which the participants were exposed to the aforementioned compression conditions in a seated position. Throughout the experience, participants were asked to ‘think aloud’ about what they felt (i.e., a usability research methodology whereby participants are asked to verbalize whatever comes to mind) [138], especially regarding compression distribution, comfort, and preferences amongst the various shoulder compression types. At the end, a qualitative interview was performed to understand the experience of wearing the garment, the compression distribution, and potential improvements to the garment.

3.1.3. Exploratory Study Results and Discussion

Since this is an exploratory study with a relatively open-ended study format that encourages users to speak their mind, the gathered data were qualitative in nature. The data was analyzed using thematic analysis, clustered to extract major patterns and themes in participant responses. The following section provides a categorized summary of participant think aloud comments, with particular emphasis on what these findings mean and how they can be used to drive future compression garment designs.

Garment Compression Distribution and Preferences

Garment compression distribution was examined since we could envision situations where uneven distribution could possibly be uncomfortable/bothersome for participants. Figure 4 shows the number of participants who felt that the compression was evenly distributed for various locations and vectors (this excludes asymmetric compression distributions across the left and right sides of the body; asymmetric distributions were disliked by all participants and corrected when encountered). Here, distribution refers to the concentration of compression sensations on certain body areas relative to others. Figure 3-4 depicts participants’ shoulder vector preferences.

Figure 3-3: Frequency count of participants’ perception of compression distribution for various compression locations and vectors for low fidelity garment

Figure 3-4: Distribution of participants’ shoulder compression vector preferences for low fidelity garment

From Figure 3-3, only 3 out of 5 (60%) of the participants felt that the compression on the torso was evenly distributed; of the two participants who thought otherwise, both were females and they felt that the bust/front chest areas were loose. This is likely due to the garment being designed for an average male. All shoulder test conditions had 4 out of 5 (80%) of participants rating the garment as evenly distributed, except ‘diagonal crossed’; only 2 out of 5 (40%) of the participants felt that the ‘diagonal crossed’ had an even compression distribution. Remarkably, ‘diagonal crossed’ was still preferred by 60% (3 out of 5) of the participants (Figure 3-4), suggesting that contrary to expectations, uniform compression distribution may not necessarily be what participants deem as their preference. This is also evidenced by one participant commenting specifically that their self-designed (informant design) strategy was unevenly distributed but it was their preference.

From Figure 3-4, participants preferred either the ‘diagonal crossed’ or the ‘straight’ test conditions; none of them liked the ‘one-side crossed’ or ‘mixed’ shoulder compression vectors. Interestingly, these two compression distributions were vastly different; ‘diagonal crossed’ was centered on the shoulder straps and ‘straight’ was centered on the chest area. A closer look into the ‘think aloud’ data found that 2 of the participants did not like the compression distribution for ‘diagonal crossed’ at all because it inhibits breathing, but 2 other participants liked the compression centered on the strap intersection because the resistance ‘feels like a hug’. During the informant design exercise, 3 of 5 of the participants wanted more compression on their backs and trunk but more flexibility on the shoulder joint, 2 participants wanted compression on their arms (this garment had no arm compression), and 1 participant felt that a stronger compression might be welcomed. We also observe that one outlier subject just did not like any form of compression on the body; this participant consistently thought the garment as uncomfortable (all conditions, including the informant design exercise) and informed us that they are typically averse to pressure on the body, and hugs are typically unwelcomed.

Although this is a small pilot study, we found evidence that counterintuitively, some participants may actually prefer an uneven distribution of compression between body locations. Further, all participants either selected ‘diagonal crossed’ or ‘straight’ shoulder compression vectors as their preference; the difference in the sensations these compression vectors provide suggest that there may be larger individual differences in terms of perceived comfort and preferences than initially expected and future designs should probe the extent of such variations.

Garment Design and Compression Comfort Perceptions

A major theme gathered in terms of garment design was that the woven canvas fabric used was too stiff—this particular woven canvas material was selected initially due to its stability. From a functional standpoint, a compliant fabric would provide more comfort to the wearer but will reduce the efficiency of any selected mechanisms in the application of compression, especially if the fabric has stretch to it. In contrast, a stiffer, non-stretch fabric would be more stable and may produce more uniform compressive forces, but a fabric that is overly stiff may not allow complex changes in interfacing with the body due to non-conformity with body contours (i.e., possibly resulting in pockets where the fabric does not touch the skin), and will require precise actuator control to create the desired compression. Hence, a clear design implication gathered would be that there needs to be a balance between compliance and stiffness of fabric, especially if/when a mechanical actuator 41 is to be integrated into the system to produce desired compression sensations.

Further, as mentioned, two female participants felt that the front chest areas were loose and remarked that the garment felt like it was designed for men (and hence was not fitting comfortably). Since it was a low-fidelity prototype and we felt that the garment afforded plenty of adjustability, we thought it would be sufficient to accommodate both genders as a quick starting point. Contrary to what was expected, the built-in adjustability still could not accommodate the gendered anatomical variances. Therefore, a clear design implication was that there should be separate designs for men and women to accommodate for varying body shapes and sizes.

In terms of the experience of compression, one of the themes gathered was the role of movement in usage of the garment. Participants commented that it was anxiety-inducing when breathing or movement is restricted, especially when the chest and top of shoulders were compressed. Several design strategies can be employed including having less compression on the outer edge of the shoulders, using a more flexible fabric to allow shoulder joint movement or torso expansion, and reconsidering placement of compressive straps.

Finally, one of the most recurring comments, made by 4 out of 5 participants (without explicitly asked), was that the length of the garment should be extended. This concerns the stark difference in compression at garment’s edge (i.e., transition between the compression-inducing garment and end of the garment interface), which was described as distracting and uncomfortable. Therefore, a more flexible interface, including the right use of fabrics, length of garment, and positioning of actuators should be considered to encourage smoother transition of compression sensations.

Physical Sensation of Compression

We also asked participants to describe how the physical sensation of compression generated by the garment compares to other forms of physical touch (e.g. a hug from a loved one)—this question was inspired by the applications seen in the literature review presented in Chapter 2 whereby many compression technologies are in fact related to mediated social touch. One participant commented that the garment was not as good as a hug because of its stiffness, while another said that the compression distribution was similar to a hug and was particularly excited about the garment since it was reminiscent of fun memories of his parents’ hugs. Several participants also requested compression to be applied not only on the torso but also the arms since ‘it would be more similar to a hug’ and might feel nice. However, two participants also commented that there were limitations 42 in the current garment because the compression provided did not include or replicate the warmth of a hug from a person. Therefore, the role of compression in garment-mediated social touch is promising but future work should look into the role of temperature in enhancing the subjective comfort and perception of on-body compression garments.

3.1.4. Exploratory Study Insights and Design Implications

From this exploratory study, the strongest theme gathered was the role of movement and breathing while compression is applied. Generally, users wanted more compression on the back, lower spine, and sides, while allowing movement (especially the arms and shoulders), and flexibility in the front for breathing. The following provides a summarized list of gathered insights and these feedbacks will be used to drive following garment development iterations.

1. Fabric stiffness greatly influenced the garment’s perceived user comfort. Compliant fabrics may provide more user perceived comfort but in turn reduce efficiency when/if actuators are integrated in the future (i.e., if a stretchy fabric is used, a portion of the displacement provided by integrated actuators will pull the stretchy fabric until the fabric reaches its stretch limits and induce a blocking force before compressive forces will be generated; hence forces generated by the actuators will not be efficiently translated). Stiff fabrics on the other hand may afford more stability and efficient translation of forces, but require more precise pressure controls since they do not conform to the body well. Hence, a clear design implication is such that fabric selection should balance between structure, compliance, functionality, and comfort.

2. One outlier subject categorically did not like any compression on the body; this participant consistently rated the garment as uncomfortable and informed us that on-body compression or hugs are typically unwelcomed. This is unsurprising since we know there exist a large spectrum of individual differences when it comes to the experience of haptic sensations on the body [16]. Yet this finding is an important one since it demonstrates that such variations in preferences also exist with compression-based haptics and we should be sensitive to the needs and wants of this group of potential users.

3. Shoulder compression preferences were divided between ‘diagonal crossed’ vs. ‘straight’ (i.e., none of the participants favored ‘mixed’ or ‘one side crossed’) and none suggested alternative forms of stimulation in the informant design exercise, which helped reduce the design space.

However, since the compression distributions between ‘straight’ and ‘diagonal cross’ were drastically different, it may indicate that there may be larger individual differences in perceived comfort of compression than expected. This is also supported by the fact that uniform compression distribution between body locations may not be what is universally preferred and hence design strategies to accommodate these differences should be considered.

4. The stark difference between presence and absence of compression stimulus at the bottom of the garment (‘edge effect’) may negatively affect user’s overall comfort. Future garment designs should encourage smoother transition between actuators. Length of the garment should also extend to the lower spine, which was a favorable location of compression for many.

5. The adjustable design did not mask gender variability in regard to garment fit. Hence, separate designs for men and women should be created to account for gendered anatomical variances.

6. Participants felt differences in compression distribution during torso movements or breathing and restricted breathing/movement may cause anxiety.

7. Arm compressions should be considered in future applications.

Technology Investigation and Selection—Shape Memory Alloys

Given the insights gathered from the work presented in Section 3.1, the next step is to consider the methods of actuator integration that will give rise to compression garments capable of delivering controllable dynamic compression inputs (such as location, intensity, duration, and pattern). Since there is currently no technology capable of delivering compression sensations in a controllable manner without the need for tethered actuation sources, while maintaining soft garment structures and elements of wearability/ usability, we pursue an engineering investigation to address this knowledge gap. In doing so, we not only add to the body of soft robotic knowledge by establishing engineering and design strategies to allow the creation of soft robotic wearable systems, but ultimately, this system enables the study of entirely new types of on-body compression experiences. The following section surveys compression technology schemes available, summarizes the advantages and disadvantages of each of the various strategies, and down-selects a promising actuation candidate for use in the next iteration of compression garment design.

3.2.1. Survey of Current Compression Technology Strategies 44

As described in Chapter 2, a variety of actuation technologies can be used to generate compression on the body. There, we discussed three most commonly used actuation strategies: pneumatics, servo motors, and shape memory alloys (SMAs). However, there are other actuator types including several classes of active materials (e.g., shape memory polymers, dielastic elastomers), haptic surfaces (e.g., deformable crusts, pin arrays), and even non-contact ultrasound transducers (e.g., midair haptics), of which the production of compression sensations on the body are theoretically possible. However, these latter set of actuation strategies are less common for various reasons, such as (1) the inability to be integrated into wearable form factors given the need for bulky supporting equipment, (2) the technology maturity at its current state does not fulfill the minimum criteria required for on-body compression generation (e.g., limited range of generated pressures/strain), or that (3) the novelty of technology, still in research stages cannot be readily utilized [139]–[142].

A brief summary of the advantages and drawbacks of each of these actuation methods are summarized in Table 3-1. For details regarding each of these actuator types, please refer to the selected references and review papers provided in the table.

Table 3-1: Evaluation of common haptic actuator types and their associated advantages and disadvantages as potential compressive actuation strategies.

Actuators Advantages Disadvantages - Fast response times - Typically require large and heavy air Pneumatic pockets/ - Precise pressure controllability compressors pouches [17], [91], - Soft components - Noisy [97], [102] - Relatively easy to integrate into fabrics - Inflation creates bulk - Fast response times - Typically require large and heavy air Pneumatic artificial - Precise pressure controllability compressors muscles [38], [143] - Relatively easy to integrate into fabric - Noisy structures Tensioning - Good existing literature on motor - Typically produces uneven pressure cords/straps with DC control distributions or servo motors [96], - Varying cords and straps materials used - Cumbersome [98], [101], [119] may offer good wearability - Still require many hard components Moving plates driven - Good existing literature on motor - Hard components by DC or servo motors control - Cumbersome [125] - Precise controllability - Noisy - Hard components - Fast response times Vibrotactile actuators - Require many actuators to produce - Small form factor (discreet & (Voice coils, linear distributed surface pressures lightweight) resonant actuators, - May produce disruptive feelings - Easy to use and integrate speakers) [44], [144], - Vibration stimuli are not easily - Good existing literature on motor and [145] scalable to low frequency / large area haptic control profiles Midair haptics [142] - Non-contacting with the user - Non-wearable form factor 45

- Lack portability - Limited range of pressures Deformable crusts and - Cumbersome pin arrays (particle - Ability to generate compression on the - Rate of actuation difficult to control jamming, swelling body surface - Takes time to return to resting state polymers) [146], [147] - Limited durability - High power consumption Dielectric elastomers - Soft components - Relatively underdeveloped as a [148] - Large active strains compression actuator - Low active forces - Irrecoverable strain effects Shape memory - Soft components - Relatively underdeveloped as a polymers [148] compression actuator - Limited contraction stroke - Small form factor (discreet & - Heat generation/ High power lightweight) Shape memory alloy consumption - Soft components wire [19], [24], [148] - One-way shape memory effect (takes - Relatively easy to integrate into clothing time and external force to return to structures resting state) - Large contraction strokes, resulting in a - Slow reaction times larger range of compression intensities - Heat generation/ high power - Small form factor (discreet & Shape memory alloy consumption lightweight) coils [35], [148], [149] - One-way shape memory effect (takes - Soft components time and external force to return to - Relatively easy to integrate into garment resting state) structures

As highlighted, the goal is to develop computer-mediated compression garments that are dynamic, low-profile, and remotely controllable. Specifically, the key innovation necessary is a garment platform that abandons traditional wearable actuation schemes (e.g. servos, hydraulics, pneumatics) and instead shifts toward the use of embedded soft robotic actuation technologies that can provide varying levels of compression sensations on the body in clothing-like form factors. Recent research in active compression garment design has focused on form-fitting garments with integrated active materials, specifically, SMAs, a type of active material that feature the ability to repeatedly change shape when actuated [133]. SMAs are promising as an actuation mechanism due to their small and soft form factors that likely satisfy a range of wearability requirements, while reducing the mechanical impedance between the device and user interface [41], as well as the possibility of easily integrating the actuators into a garment structure. Their well-documented performance measures as actuators capable of providing significant forces and large active displacements when in select configurations (hence can be used to dynamically apply compression on the body), makes them a particularly attractive actuator option. Further, the use of SMAs for compression applications is understudied (compared to pneumatic actuation mechanisms), especially for upper-

46 body garments involving large body areas (currently most SMA-based technologies are limited to single/localized body regions [24], [35], [75]), and therefore worth investigating as an alternative, non-traditional actuation scheme. The following section describes the various mechanisms of action, features, and characteristics afforded by SMA actuators, as well as how this class of actuators could be used as a method of applying compression on the body.

3.2.2. Shape Memory Alloys (SMAs)

SMAs are a category of active material that demonstrates shape memory effect, which can be described as the ability to return from a deformed state, to a prior pre-deformed, ‘remembered’ state upon exposure to a thermal stimulus. This shape memory effect happens as a result of the alloy undergoing solid-state transformation between a low and high temperature phase, and it can take one of three possible crystal structures: twinned martensite, detwinned martensite, and austenite [150]. Heating an SMA above its transformation temperature triggers a transformation cycle from the low temperature martensite phase to a high temperature austenite phase. SMAs also demonstrate superelasticity behaviors [151], which points to SMAs’ ability to recover relatively large strains (up to ~6-10%) upon mechanical loading-unloading, where the atomic bonds in the materials’ atoms stretch to extreme lengths without inducing permanent deformations [151], [152]. This response can be modified in a variety of ways including altering the composition of the SMAs to tailor specific critical temperatures, or heat treating the material at specific (typically high) temperatures to achieve a desired activation shape [123].

There are various types of SMA alloy compositions; the two main types being copper-aluminum- nickel (CuAlNi) and nickel-titanium (NiTi). For the purposes of this project, only NiTi SMAs (commonly known as Nitinol or Flexinol) are used since it is commercially available, biocompatible, and typically used in many applications such as medical devices (e.g. stents [153], catheters [154], orthopedic implants [155]), robotic system actuators [156], [157], and compression systems [158]. SMAs are typically manufactured and used in a wire configuration. However, in a wire form, SMAs are limited in terms of their force output due to small active strains of only 6-10% (i.e., it does not produce much linear displacement) [151], [152]. This is evidenced by several past wearable compression applications with integrated SMA-wires, and found the compression forces generated to be too weak for it to be effective [18], [19], [151]. For the purposes of applying compression on the body, the material’s shape memory effect is best exploited by forming the SMA wires into coiled, spring-like actuators to amplify their active/recoverable strains by an order of 47 magnitude (i.e., upwards of 50-75% recoverable strains are possible in this configuration). There are various ways to manufacture and post-treat SMAs based on their intended target applications, for a detailed explanation of the various methodologies and alternatives, refer to the work by Rao, Srinivasa, & Reddy [151] and Kumar & Lagoudas [159]. For the actuators used in this project, first, raw NiTi SMA wire was wound into a tightly-packed coil using the method in Figure 3-5 [160]. Next, the coils were heat treated at high temperatures to set their shape (i.e., trained to retain their tightly packed coil configuration) [161]. After which the coils were then cooled in water, which sets their recoverable memory shape in the form of a tightly packed, fully contracted spring. After the series of SMA coil shape setting steps, the SMA spring actuators are ready to be used in a 3- step activation cycle, described in Figure 3-6.

Figure 3-5: (Left) Method of producing tightly packed coils from SMA wire. (Right) Comparison of tightly packed coils vs. SMA coil that has been physically stretched out. Obtained from [160]

Figure 3-6: SMA coil actuator shape setting steps and activation cycle. Figure obtained from [160], [161]

The shape-set, twinned martensite actuator is the beginning of the cycle (Step 1). Upon deformation by an external force at low (in this case, room) temperatures, the coil actuator stretches (i.e., de- twinned in martensite phase) (Step 2). Subsequent heating of the stretched actuator triggers the austenite phase change, causing contraction of the actuator, which produces active forces and displacement (Step 3). These SMA spring actuators that constrict from a deformed (i.e., stretched out coil structure) spring configuration, into a tightly packed (i.e., contracted coil structure) spring configuration when heated, have been shown to produce controllable compression (>225mmHg) that scales with applied current [123]. For this dissertation, all SMA actuators are manufactured with Flexinol® wire (SMA wire diameter= 0.012”, activating temperature= 70°C) into 0.048” outer diameter springs (spring index= 3, i.e., ratio of spring diameter to wire diameter, selected according to [160], [161]), and heat-treated at 450°C for 10 minutes to set their shape.

SMA coil actuators are particularly appealing as a solution in this context for a variety of reasons. The first being their large active displacements (up to 4:1 recoverable extension/contraction ratio), which makes them suitable as a mechanism of donning and doffing (in a loose garment form when the SMA coils are stretched out), without compromising their ability to dynamically apply compression as needed (when the SMA coils contract). Further, prior work has shown their active forces can be used to augment compression produced by passive garments; the coupling of passive fabric and active SMAs is capable of producing compression around a body part [123]. This concept is illustrated in Figure 3-7. In terms of comfort and discreetness, both of which are important parameters in a successful wearable solution, the small form factor of long, small diameter elements (coil outer diameter= 0.048”) is conducive for integration into textile structures. 49

In this simplified model (in the ideal case where the combined SMA actuator-fabric length is equivalent to the circumference of the body area), SMAs (twinned martensite) are attached in series with a passive fabric strip (Step 0). As the fabric and SMAs are donned circumferentially around a body part, they form a passive equilibrium state, XP with an initial passive compression (Step 1). Here, the SMAs are fully stretched out and in their de-twinned martensite state (i.e., experiencing high strain). As the SMA actuators are thermally activated (Step 2), the SMAs contract (austenite phase), pulling/further stretching the passive fabric, which produces increased tension that imparts a compression sensation to the underlying body region. Here, a new active equilibrium state, XA is achieved (Figure 3-8). In other words, as illustrated in the force-length relationship diagram, changes in SMA length (contraction) causing equal and opposite passive fabric length changes (extension) result in an increase in system force and thus, increased compression sensations.

Figure 3-7: Simplified compression system activation process, with SMA and passive fabric coupling, as well as representative force-length relationship for the described SMA-based compression system. Figure adapted from [123].

Figure 3-8: Representative illustration of the SMA-actuated compression system wrapped circumferentially around a body part

Perhaps more importantly, these SMA coil actuators can be conveniently activated by an applied voltage (or even potentially though body heat provided the activation temperatures are sufficiently low [162]), and has the potential to provide controllable compression since its actuation performance scales with applied power. Previous work by Holschuh et al. has also shown that it is possible to model and predict the passive (Eq 1) and active (Eq 2) pressures applied through a garment integrated with SMA coil actuators, based on the following model and system parameters (refer to [123] for equation derivations and Table 3-2 for variable definitions):

퐸푡 2휋푟−퐿푆0 (1+ ∈푆푚푎푥) 푃푃푎푠푠푖푣푒 = ( ) ( − 1) (Eq1) 푟 LF0

2 ∆푋푠푦푠푡푒푚퐺퐴푑 푛푎퐸푡 푃퐴푐푡푖푣푒 = 2 3 (Eq 2) 푟(퐺퐴푑 푛푎퐿퐹0+퐸푤푡8퐶 휂퐿푆0)

Table 3-2: Variable definitions for predicting active pressure applied through SMA-integrated garment Parameter Description Units Source of Measure C SMA spring index = D/d - Actuator manufacturing process d SMA wire diameter m Manufacturer specifications D Coil diameter m Actuator manufacturing process E Passive fabric Young’s modulus Pa Garment manufacturing process

GA SMA austenite shear modulus Pa Prior literature

LF0 Passive fabric unstretched length m Garment manufacturing process

LS0 SMA twinned martensite length m Garment manufacturing process

na Number of parallel actuators in a system - Garment prototype manufacturing process r Limb radius m Use scenario t Passive fabric material thickness m Garment prototype manufacturing process 51

w Passive fabric axial width m Garment prototype manufacturing process

∈Smax Maximum SMA de-twinned extensional - Use scenario strain

ΔXsystem Unstretched system closure gap= (2πr – other m Garment prototype manufacturing non-fabric structures- (LS0 + LF0)) process η SMA spring packing density - Prior literature

Using the model above, several design insights that can be extracted that will be helpful to consider when designing the SMA-integrated compression garments. The first is to minimize passive pressures—with decreased passive pressures, experientially, the donning/doffing process will be easier on a user. Functionally, from a sensory stimulation perspective, the relative impact of active pressures will be maximized with lower passive pressures (i.e., if the garment already imparts high passive pressures, additional pressures imparted with SMA actuation may not be easily perceived; more formally explained by Weber-Fechner Law [55]). However, if there is no/too low initial pre- tension in the passive fabric, the total active force that can be achieved will be lower. This is because a portion of the actuator displacement will then have to be used to bring the system to its passive equilibrium state. Hence there needs to be a balance; a certain amount of passive tension in the fabric is necessary but should not impart such high pre-tensions such that additional compressions can no longer be detected. To minimize over-imparting passive pressures (assuming the targeted body region radius is fixed), the passive fabric unstretched length (LF0) and SMA twinned martensite length (LS0) could be maximized or alternative strategies such as decreasing passive fabric Young’s modulus—a measure of the tensile stiffness of a material (E), and passive fabric material thickness (t) could be employed [123].

Secondly, is the method to maximize active pressures—this is of interest to this dissertation’s research since the goal is to impart compression sensations on a user and we are interested in a range of compression sensations. From Eq 2, we note several variables that can be controlled to increase active pressures. Since the SMA actuator manufacturing process has been selected and is not within the scope of this project, the variables that can be easily manipulated in the garment construction process to maximize compression sensations include the number of parallel actuators in a system (na), passive fabric Young’s modulus (E), and passive fabric material thickness (t). This means that if a stiffer passive material is selected, the system generated forces will be higher than those generated by a less stiff or stretchy fabric [123]. However, from the feedback gathered from the pilot study described in Section 3.1, this could be at the expense of user perceived comfort and/or increased passive pressures as described earlier, i.e., a softer fabric with more compliance 52 will be more comfortable for users but result in lower generated compression. As a practical design implication, a compliant, stretchy fabric (that is likely to satisfy comfort requirements) will overstretch and allow the SMA actuators to contract too much so that the equilibrium tension is unsatisfactory low (which will likely not satisfy functional compression requirements). Referencing Eq 2 in the context of Figure 3-7, with the fabric’s Young’s modulus (E) as the slope of the fabric’s theoretical curve, a shallower fabric modulus curve will produce lower equilibrium tensions, longer fabric extensions, ultimately resulting in greater SMA contraction-displacements before arriving at adequate compression levels.

Further, an increased number of actuators presented in parallel in a system is also another method of increasing active pressures, however, this will in turn be at the expense of higher power consumption and thermal generation, potentially impacting system efficiency and user experience. Hence, while the above model can help guide design decisions, the relative drawbacks and potential impact of the engineering strategies on user comfort has to be experimentally determined through user studies to strike a balance between technical feasibility and overall user experience. For detailed explanation on the relationship between the variables supplemented with empirical data and what that means in terms of the range of flexibility in design refer to [123], [160].

While the above model provides guidance for SMA-based compression garment design decisions, it should be noted that this model is based on several assumptions: (1) pressure is generated on a rigid and cylindrical object, (2) negligible friction, (3) uniform circumferential tension application, (4) SMA and passive fabrics both operate in the linear region, (i.e., show linear stress-strain behavior), hence can be modeled as linear springs, and (5) passive fabric thickness remains constant during stretching [123]. This model also only accounts for predictions during full SMA transformation, therefore, it only predicts the maximum pressure generated and not intermediate actuation levels. All of these assumptions while simplifying the model, affects its overall accuracy and predictive power.

This brief discussion about SMAs will not be complete without pointing out some of the limitations and potential drawbacks of SMA technology. While SMAs can provide large actuation forces with a simple electrical setup (i.e., running current directly through each spring to induce resistive heating), they have relatively low energy efficiencies that may result in high power consumption without adequate design strategies. Further, since the transformations in alloys are thermally- induced, the thermo-mechanical behaviors are difficult to model accurately (per the limitations in 53 the mathematical model presented by Holschuh [160]), especially when used in complex geometrical surfaces like the human body; all of which will result in complex control schemes if a feedback loop is intended.

Besides, the SMA manufacturing methods presented here results in actuators with one-way shape memory effect; so when the alloy cools into its twinned martensite form, it will retain that shape until physically deformed again. In other words, when integrated into a garment, the material can induce a compression force but is unable to ‘un-compress’ (i.e., relax) unless the actuators are physically strained by an external source (i.e., the coils are manually re-stretched) or separate design strategies are integrated (this work will present a design strategy to overcome this later in the chapter). Also, the temperature dependent effect of SMAs also result in the actuators typically having lower reaction times compared to other conventional actuation mechanisms since heat accumulation is needed to cross the alloy’s transformation temperature threshold. This also means that these actuators (unless adequate design strategies are implemented), should not be used for applications requiring high frequency haptic presentations. This should not present a problem in this current compression application since it was found that people generally prefer a slower ramp up of compression when presented on the body; it was seen as more pleasant and comforting [22], [30], [163] (i.e., the slower reaction times is unlikely to be a huge hindrance).

In summary, SMA coil actuators are a particularly appealing solution as a method for providing compression since they can provide large active displacements that can be used as a mechanism of donning and doffing loose garments while retaining the ability to dynamically apply compression, and can be easily and unobtrusively integrated into typical textile structures (but their limitations should be considered for optimal use of the technology). With that understanding, the following sections will describe the process and lessons learned given the coupling SMA actuators into textile structures to develop technologies that are capable of delivering dynamic compression on the body.

Feasibility Study with Medium Fidelity Prototype

Motivated by the various advantages associated with SMAs, we sought to determine ways to integrate SMA actuators into a garment that can produce perceivable compression sensations on the body while preserving other wearability considerations (e.g., minimal mass/bulk/rigidity, reduced heat exposure), as well as understand how people respond to this form of dynamic compression—this is the first instance of the use of SMA actuators for compression application on 54 large areas of the body. The following sections detail the engineering and design process to understand tradeoffs in system performance based on differing design parameters—in doing so, we settled on a series of intentional design choices that collectively, enable the compression functionality that is required. It is also our intention to preserve these intentional design choices and user feedback to help future designers to operate in this research space. The new system design, henceforth referred to as ‘medium fidelity prototype’, was advanced based on feedback informed by users from the low-fidelity prototype user study [164].

3.3.1. Medium Fidelity Test Garment Design

This medium fidelity dynamic compression garment prototype utilizes nickel titanium (NiTi) SMA coils integrated into the textile structures to produce compression on the body. The garment went through several prototype iterations to understand actuator placements, fabric use, sizing, and fit. Here, we briefly present some of the key findings. For instance, we took inspiration from the design by Duvall et al. [165] that integrated SMAs by looping a single actuator around fiberglass tubes in a ‘snake-like pattern’ (Figure 3-9a) (i.e., electrically, the actuators were connected in series). However, we found that the developed garment was extremely stiff, and due to the serial SMAs configuration, resulted in uneven compression applied (the SMAs were caught in the fiberglass tubes and the single, long SMA length resulted in uneven electrical heating, which did not facilitate a controllable/predictable compression behavior). Also motivated by feedback in the low-fidelity garment pilot study, we used softer, more compliant fabrics (Figure 3-9b) with the SMAs connected in parallel. However, the slightly stretchy fabric resulted in the SMAs stretching out the fabric instead of applying compression (as explained in the previous Section 3.2.2 on theoretical modeling of the system), which means the actuator efficiency was reduced and low-to-no compression was felt. Further, the actuator length in Figure 3-9b was insufficient to produce discernable compression sensations. Upon further iterations and bench-tests, we landed on the final design in Figure 3-10.

Figure 3-9: Prototype iterations of SMA-integrated garments. (a) Left: SMA connected in series; (b) Right: SMAs in parallel in conjunction with softer fabrics

The garment (Figure 3-10) consists of an inner comfort layer and an outer actuation layer, with both layers connected through a zipper (for ease of donning and doffing). The inner comfort layer was created through a combination of woven canvas fabric on the front, 3D spacer foam-mesh on the back (for ventilation, body-contouring/back support, and comfort), and the heat insulation side panels (since the SMAs actuate using heat) positioned directly below the SMA coil actuators. Each of the heat insulation panels further consists of 3 sub-layers: cotton aramid (compliant base comfort layer and a flame retardant), reflective heat shield (middle layer, for heat management), and polytetrafluorethylene (TeflonTM) (top layer, low friction surface for smooth actuation). The outer actuation layer consists of non-stretch cotton aramid, with SMA spring actuators evenly spaced 1.0-1.5 inches apart, connected to the actuation layer using metal snaps in a parallel circuit configuration. Using between 4W-18W (depending on the number of actuators in the circuit) of power from a DC power supply, the SMA springs actuated on the order of 3-10 seconds. 56

Seven SMA actuators were used on each side on the torso, five actuators were used on the armbands, and three actuators on the shoulder regions. Metal snap connectors were used to not only form stable electrical connections, but also to enable the actuator positions in the shoulder regions to be changed (as an extension of the low-fidelity study, different compression vectors can be applied and studied). Fiberglass tape was placed down the side seams to mechanically and electrically separate each actuator. Following participant feedback from the low fidelity garment, the medium- fidelity garment included more compliant fabrics, extended garment length, actuators placed away from garment edges, added armbands without restricting shoulder joints, and lowered the neckline. The garment was constructed to a standard male size S with a 36-inch chest circumference.

3.3.2. Medium Fidelity Prototype User Study Methodology

The constructed garment was tested with a small user study group; the goal was (1) to evaluate the functionality of the system—if the compression sensations provided were felt by users, and (2) to study how users respond to this form of SMA-based compression application since this form of SMA, soft robotic integrated soft goods (upper body) garment has yet to be tested on participants. The medium fidelity SMA garment study used a within-subjects design with 8 healthy male participants (age range= 18-27, average age= 23.5). To prevent issues with sizing and fit as previously noted, only self-identified size S male participants were recruited. During the study, each participant wore a cotton long sleeved T-shirt to ensure equal baseline conditions. The testing room temperature ranged 19-22°C.

Each participant was exposed to 5 compression conditions: ‘torso’, ‘shoulder-straight’, ‘shoulder- diagonal’, ‘shoulder-mixed’, and ‘shoulder preference + arms’. All shoulder conditions Figure 3-12 a–c) included torso actuation. The ‘straight’, ‘diagonal’, and ‘mixed’ shoulder compression vectors were down selected from the exploratory pilot study with the aim of further narrowing the design space and investigate how users respond to these varying shoulder compression vectors provided through SMA actuation. The study was pseudo-randomized; the ‘torso’ condition was performed first because the shoulder test conditions are contingent upon anchoring of the trunk to prevent the garment from riding up as shoulder SMAs are actuated. The ‘shoulder preference + arms’ condition was investigated last; it uses the most preferred shoulder condition with armbands, and the three shoulder test conditions were randomized. The varying compression locations were investigated as such (i.e., instead of having a single compression intensity levels for all body locations) since the low fidelity exploratory study has demonstrated that uniform compression distribution between body locations may not be what is universally preferred.

Figure 3-12: Medium fidelity garment shoulder test conditions (a) ‘Straight’ (b) ‘Diagonal’, (c) ‘Mixed’ shoulder compression vectors

For each test condition, the participants donned the loose-fit, unactuated garment and the actuators were activated to provide compression for 1.5 minutes. The participants sat with both their hands on the table to provide consistency in postures across test conditions and to prevent the underside of their arms from coming into direct contact with the SMAs (actuators were uninsulated for the medium fidelity prototype). During each test condition, the participants completed a survey about comfort and distribution of temperature and compression on the body, and they were also prompted to ‘think aloud’ throughout the testing period, especially on the comfort of compression and temperature on the body. Qualitative interviews were also conducted to understand the experience of wearing the garment, the compression and temperature distribution, emotional changes while compression is applied, interpretations of the compression, and potential improvements to the garment. As before, the data was analyzed using thematic analysis and major patterns and themes were extracted from participant responses. 58

3.3.3. Feasibility Study Results and Discussion

The following section provides a categorized summary of participant comments, with emphasis on how participants feel about the experienced compression sensations provided by the SMA-driven garment, as well as gathering engineering and design insights to be used to improve and drive future compression garment designs.

Garment Overall Preference and Rating

After all test conditions were completed, the participants were asked to provide explicit test condition preferences and to rate the compression garment on a scale of 1 (extreme dislike) to 10 (extreme liking). 6 out of 8 (75%) participants selected the ‘straight shoulders’ as their favorite shoulder compression condition, while remaining participants selected ‘diagonal’ shoulder condition as their favorite (Figure 3-13).

Figure 3-13: Participants’ shoulder actuation vector preference for medium fidelity prototype

Consistent with the exploratory study, none of the participants preferred the ‘mixed’ shoulder test condition. Through the participants’ think-aloud comments, the reasons for disliking the ‘diagonal’ condition was the feeling of a strong pressure on the collarbones due to the actuators seated on top of the area, as well as the actuators being too close to the neck area and the heat could be felt. The mixed test conditions had the most complaints of feeling too strong of a pressure on their armscye (‘arm holes felt too tight’) as well as being difficult to breathe. As mentioned, the armband preference was split between liking and disliking; some felt it had little value and was restrictive, while others were very satisfied. Overall, the average liking rating for the garment was an 8.4/10.

Garment comfort

Overall, 6 of 8 participants said that the garment was comfortable. All but one felt that the garment was well-fitted; this participant was of smaller stature and the garment was too long which caused gaping in the shoulders. One participant even commented that the garment felt like a ‘second skin’ and he had not had a compression garment that is so fitted. We noted the importance of garment fit since it affects the compression profile provided by the garment. From the low fidelity garment pilot study, we learned of the importance of allowing mobility and therefore a major design change was the use of more compliant non-stretch fabric. Interestingly, we discovered an unexpected advantage with the SMA garment—when activated, the garment provides mechanical resistance to produce compression but can still accommodate participants’ movements such as cyclic torso expansion while breathing. This is possible because when the SMAs contract, they pull the passive fabric close (i.e., the coils are tightly packed), providing a certain amount of resistance to produce compression. When a user moves/breathes, a load is applied and the spring extends (i.e., the coils are stretched out), allowing for the accommodation of the induced movement while remaining in tension. In other words, the actuated garment is tight, but is remains stretchy, so user movements are accommodated; a participant even clearly pointed out (unprompted) that the garment mobility was much better after the SMAs were actuated compared to the unactuated garment.

Some participants felt that there was a gradual compression starting from the torso to the shoulders; future iterations will synchronize the rate of actuation on various locations. In terms of compression distribution, there were a few instances where participants felt that the compression on one side was stronger than the other, and a heavier concentration of compression relative to other areas (i.e., pressure hotspots) were deemed uncomfortable. This uneven distribution may be due to a variety of factors including slight differences in power input or actuator lengths. While not all participants noted this uneven loading, future iterations should pay particular attention to potential pitfalls in this area, including calibrating the garment actuation such that it provides symmetrical compression and/or better characterize the pressures applied.

In terms of temperature distribution, some participants reported that the garment started to feel too warm at times, experiencing rapid breathing during such encounters. Therefore, for the next garment iteration, efforts will be made to evaluate materials that provide better thermal insulation. Also, interestingly, half of the participants commented that the lower back felt warm, even though the lumbar area did not have any integrated actuators. This might be due to the back and spine being 60 a high heat and sweat zone [166]. While we anticipated the need for ventilation and used a foam- mesh material for the back, the problem of back heat management remains evident. An implication for future design includes the need for a more thorough consideration of thermal sweat profiles to better modulate the thermal comfort on given body areas.

Emotional Changes, Comparison to Hugs, and Envisioning Garment Usage

As with the low fidelity passive garment study, the interview also probed the participants’ subjective emotions and how the physical sensation of compression generated by the garment compared to other forms of physical touch (e.g., a hug)—we anticipated varying responses due to the fact that this garment has computer-mediated technology integrated and will likely influence participants’ perceptions. Two participants commented that ‘something moving on the body’ as being weird yet fun since it was a new sensation, with a few using the word ‘exciting’ to describe the experience due to the novelty of technology. One participant also demonstrated particular excitement for the condition with the armbands, stating that when the arms were compressed, he felt more active and empowered, ‘like I’m getting ready to exercise’. On the other hand, 5 of 8 participants claimed that wearing the garment felt calming and/or relaxing to them but only when the compression and temperature distribution were even and consistent. One participant described the sense of calmness descending given the sensation of compression on the chest especially when taking a breath. Some compared the garment to the sensation of cuddling and that the ‘front felt like a hug but the back like a massage chair’. Another participant compared the garment to a heavy- weighted blanket that he uses; but the garment was said to be less suffocating and much more form fitting (unlike blankets that just provide a sensation on the top of the body area it covers).

However, 3 of 8 participants pointed out that, while the physical sensation of the garment might be similar to a hug, it is missing the emotional component of real hugs (e.g., joy, love) and feels artificial. The participants were also asked about their subjective feelings after garment removal. While 2 participants did not report any changes, others reported lingering warmth and compression. Interestingly, 4 out of 8 of the participants commented on wanting the garment back on immediately after removal without an explicit question asked. All these subjective emotional effects warrant future studies to clearly understand the positive/negative sensations associated with applied on- body compression. We were also interested both in how people might respond to computer- mediated compression as a stimulus, and also the ways it might be interpreted. With that, we asked participants to describe potential situations they might expect/envision to use the garment. The 61 most common association was athletic applications (5 subjects), followed by calming, de-stressing therapeutic devices (3 subjects), and medical purposes for muscle soreness/pain (2 subjects).

3.3.4. Feasibility Study Insights and Design Implications

This medium-fidelity prototype study showed that the use of integrated SMAs to selectively apply compression on the body was indeed possible; we not only included participant reactions to this form of soft robotic-actuated upper body garments, but also synthesized the design lessons learned throughout the process. Generally, the participants were able to detect the applied compression and distinguish between shoulder compression locations, shown in their distinct preference and ‘think aloud’ comments. The subjects generally welcomed compression on the torso and that the preferences for shoulder regions are dependent on the direction compression is applied. In terms of how people interpreted and responded to the compression stimulus, we found generally positive feedback on comfort and emotional effects. Some common themes involve calming/comforting sensations and is suggested to be potentially useful in health contexts (exercise or stress-relief). These preliminary results show that there is potential in the use of distributed compression on large areas of the body as an interaction modality, since it is detectable, welcomed if done correctly, and can be induced within a low-profile garment form factor. The following provides a summarized list of gathered design insights and these feedbacks will be used to drive the design process for future garment developments.

(1) SMA actuators improved garment mobility—We witnessed the unexpected advantage that SMAs provided in this medium fidelity garment—improved mobility. When activated, SMAs contract, pulling the passive fabric close; the garment provides enough resistance to produce compression, but when the user breathes/moves, the spring actuators can extend while still providing compression. Hence, delivering the impression that mobility was improved when the SMA actuators were activated. This justifies the use of SMAs as an actuation mechanism to apply compression on the body.

(2) Thermal considerations of SMA actuators—Since SMAs activate through Joule heating, as such, many subjects felt concentrated warmth in the lower back (even though SMAs were not present there), and at times the warmth sensations were said to be overwhelming when used for an extended period of time. Since this is a medium fidelity prototype and the first time SMAs were integrated into the garment, less focus was placed on insulating the SMA actuators, which may have 62 inadvertently caused some of these temperature perception artifacts. However, since SMAs actuate with heat, this presents an inherent limitation in the technology and has to be acknowledged—the warmth and compression may not be entirely dissociable from one another. Designers and researchers using these SMA-type technologies should be cognizant of the pros and cons the additional thermal stimulus brings. On one hand, the thermal stimulus may be distracting or even masking the compression stimulus if not handled correctly, but on the other hand, the thermal stimulus may contribute to the pleasant sensations that accompanies the compression stimulus. As per the low fidelity prototype study, some participants felt compression alone did not adequately replicate the sensations of a hug because the thermal ‘warm’ components of human touch were missing. In the feasibility study with SMAs, participants alluded that the physical sensations are similar to a hug. When viewed from this lens, the additional thermal stimulus from SMA activation can be seen as a distinct advantage compared to other types of materials in simulating human touch. Regardless, with better design strategies, it is believed that a balance can be achieved; either with evaluating actuators that have lower activation temperature profiles, insulating the actuators with materials that provide better thermal insulation, or integrating power modulation strategies (e.g., pulse width modulation) to make this actuator type more wearable for long-term applications.

(3) Down-selecting shoulder compression vectors to ‘straight’ and ‘diagonal’—In terms of down- selecting compression vectors, we gathered that the ‘straight’ shoulder condition was the most preferred (6 of 8 subjects), and the remaining two participants favored the ‘diagonal’ shoulder compression vector. The ‘mixed’ condition was not preferred by any participants, citing too strong of pressure on the armscye and difficulty breathing (chest compressed). This, combined with the low fidelity exploratory pilot study results, provides more support for down selecting compression vector designs to only include ‘straight’ and ‘diagonal’ shoulder compression vectors.

(4) No conclusions can be drawn from arm compressions—From the low fidelity prototype exploratory study that reveal several participants eager to try out arm compressions, we found from this study that arm compression preferences were equally split; some felt it had little value and was restrictive (n=4), while others were very satisfied (n=4). Currently, no conclusions can be drawn about arm compression and should be further investigated.

Considering the gathered insights from the two low and medium fidelity prototype studies, the major findings directly pertaining to garment comfort include: (1) the need for mobility (to allow joint movement and breathing), which we were pleased to discover the SMA actuators actually 63 facilitated, and (2) the need for uniform stimulus distribution (any hotspots are disliked). It should be clarified that our observations indicate that uneven distribution between body locations seem to be tolerated or even welcomed at times (according to the low fidelity exploratory study); for example, a stronger compression sensation on the shoulders compared to the torso may be what is preferred by certain users. What is disliked by participants in terms of compression and temperature distribution, are stimuli hotspots or asymmetrical distributions. A clear design implication is that future designs should consider separate controls for different body locations (possibly allowing customization of settings) but ensure the maintenance of a uniform distribution within the distributed body region and ensure that the haptic stimuli presented are symmetric between both right and left sides of the body. Further, given this was still a preliminary study, the participant pool was limited to only include size S males. A separate women’s garment will be designed to capture the variances in experience and to draw a complete set of design principles. Future work will include not only all genders, but an increased number of participants, such that the findings can be better generalized. Through that, we also hope to better understand the important problem of adapting this garment form factor to diverse body types.

With the medium fidelity prototype, we demonstrated the ways SMAs can be used to provide compression on the body and the variable control of compression on different body locations. From this, we also captured initial user reactions to the stimulation to inform future SMA-based compression garment designs; more detailed user experience studies should be implemented. A first step is to integrate control schemes to vary the garment’s compression presentation, which may include varying levels of compression intensity, location, duration, and pattern. With these controls integrated, this garment can be used to further study the user experiences of compression on the body such as identifying compression perception, subjective effects of variable magnitude/ timing/location of compression, and emotional effects stemming from this form of haptic stimuli.

One of the central motivations of integrating SMA actuators into a fabric structure was to ensure elements of wearability is maintained. However, with this medium-fidelity prototype, the garment was still tethered to a large DC power supply and manually controlled through direct connection to a power supply to provide compression; this of course limits the use potential of the system. However, this system could be easily computerized in the next prototype iteration, including both on-board power systems and remote actuation capabilities to enable an actively controlled, computer-mediated SMA-based garment architecture.

High Fidelity Garment Prototype Design Iteration

Section 3.3 demonstrated the use of SMA coil actuators in compression garments, as tested on a small group of male users. Given the newly formed understanding of the design space, we consider the following design requirements for an ideal active compression garment system:

1. Soft-goods-based system capable of controllable dynamic compression inputs (including compression location, intensity, duration, and pattern) at varying levels. 2. Fully enclosed system with on-board power and system controls to afford enhanced usability and portability. 3. Equipped with remote actuation capabilities. 4. Unobtrusive form factor with minimal bulk from actuation/control components. 5. Strategic fastening design for easy donning/doffing. 6. Socially acceptable aesthetics and discreet form factor for daily use.

The medium-fidelity garment presented in Section 3.3 only addresses a small portion of the above design requirements; it could only provide a single squeeze intensity, was tethered to a power supply, used inflexible Teflon materials, and supported only a single user size/gender. To address these limitations, further garment iterations were investigated, and the advanced prototype is presented herein, henceforth referred to as high-fidelity prototype [167], [168]. Several new engineering solutions were identified to make this technology both more functional (from a compression deliverance and relief perspective) and wearable (e.g., design strategies to further streamline form factor, reduce stiffness/bulk, and heat)—the design choices are detailed as below.

The upgraded system consists of an inner comfort layer, middle actuation layer, and outer covering (Figure 3-14). The inner two layers are connected through a front zipper for donning/doffing (Figure 3-14A). The inner comfort layer consists of woven canvas on the front, spacer foam mesh with ventilation channels on the back (for heat management/back support) and insulation panels positioned directly below the SMAs (to protect users from heat generated). The middle layer consists of non-stretch cotton/aramid, with actuators connected to actuation layer with snaps. The fabric materials used in this garment iteration was more compliant than those used in the medium fidelity prototype (e.g., the use of Teflon was eliminated; it was selected initially to provide a low- friction surface for the actuators to operate on but with the SMA form factor in this improved garment, it was found to be unnecessary). The outer covering (Figure 3-14B) is graded slightly larger to prevent interference with SMA actuation and includes pockets to house electronics. The system supports size S males and females.

As before, the garment system utilizes NiTi SMA springs (made from Flexinol® wire) to produce compression on the torso, shoulders, and arms. A total of 12 (female version due to shorter garment length) or 13 (male version) actuators are located on each torso side, 7 on each armband, and 5 on each shoulder. For shoulders, the compression vectors were further down-selected from the previous medium-fidelity study to only include ‘straight’ (oriented vertically) and ‘diagonal’ (towards chest). The actuators are spaced ¾” apart on torso and arms and ½” apart on the shoulders. Metal snap connectors were used to form stable electrical connections between actuators and fabric and to enable re-alignment of actuators as desired to create a variety of spatial compression profiles.

In this iteration, the major improvement made is the use of a braided outer sheath (¼” Techflex Flexo) for each SMA actuator (Figure 3-15).

This SMA-braided sheath implementation was done for a variety of reasons:

(1) Electrical isolation—Because the SMAs activate through Joule heating via an applied current, the actuators should be electrically isolated to prevent contact between the actuators (otherwise they may short circuit), or with the users.

(2) Heat management—The use of thermal energy to activate the actuators mean that heat will be generated during the SMA-driven compression activation. In the medium fidelity prototype user study presented previously, some participants commented on feeling overwhelmed by the heat intensity on certain body areas when exposed to the SMA actuators for extended periods of time. Therefore, heat management solutions to reduce the impact of heat transfer from the SMAs to the users, such as these insulative braids, were employed. From a safety perspective, this is also to prevent users from coming into direct contact with the warm actuators (which may cause burns if users are in direct and/or prolonged contact when the actuators are activated at their maximum).

(3) Overcoming SMAs’ one-way shape memory effect—Compressing the braided sheath causes elastic energy to be stored (analogous to compressing a spring), which forms a naturally antagonistic system that acts to forcibly re-stretch the SMA coils when unpowered. In other words, when the SMAs are powered, the braid-SMA system contract to provide compression sensations (the passive braids contract as well due to the constricting forces presented by the SMAs). When power is shut off, the braids have a tendency to elongate to its original uncompressed length so the SMAs are ‘deformed’ to a relaxed position with the elastic energy stored in the braid during SMA contraction. Hence, achieving the contraction-relaxation cycle needed. Figure 3-16 demonstrates the force output (7N) and relaxation behavior provided by a single SMA-braid actuator system.

To afford varying compression intensities of low/medium/high, a strategy of selectively activating alternating parallel actuators was employed (i.e., changing the number of actuators activated depending on the desired compression levels). For each body location, the actuators were distributed into 3 independently-controllable channels, enabling differing compression intensities through selective actuation of 1, 2, or all 3 channels, respectively (Figure 3-17 and Figure 3-18). In other words, more SMA acutators were activated with increasing intensity requirements. This in turn, increases the compression intensity levels due to higher fabric tension, as discussed in Section 3.2 (supported by the models based off the the adapted hoop stress formula, as well as empirical data demonstrated in the work of Holshuh et al. [123], [161]). The system weighed 1.2-1.4 kg without batteries (2.35 kg with batteries).

SMA Actuator Channels* Body Male Female Location Channel Garment Garment A 3, 7, 11 2, 6, 10 B 1, 5, 9, 13 1, 4, 8, 11 Torso 2, 4, 6, 8, C 3, 5, 7, 9, 12 10, 12 A 1, 5 1, 5 Shoulders B 3 3 C 2, 4 2, 4 A 2, 6 2, 6 Arm B 4 4 C 1, 3, 7 1, 3, 7

Low Medium High

The garment can be powered by 4 rechargeable LiPo batteries (Tenergy 7.4V,6000mAh, 5A), enabling untethered (mobile) operation. To afford remote-controlling, Bluetooth modules (HC-05) and electronics were connected serially to an Arduino Mega (to parse/relay Bluetooth or digital signals). To activate the garment, current must pass through the SMA network to trigger actuation via Joule heating. The networks are MOSFETs-driven (Vishay Siliconix, N-CH 30V 6A) via Bluetooth from a computer, and the current flowing through each actuator network is fine-tuned using potentiometers to ensure that each SMA-branch receives approximately the same amount of current (TT Electronics, ½” diameter); this was done to minimize asymmetry in actuation (due to slightly different actuator lengths and hence power requirements) as described by the medium- fidelity user studies. A user interface is created with Processing to wirelessly control compression location/intensity/timing is presented Figure 3-19.

Each SMA actuator received ~0.3A of current and actuated on the order of 2-8 seconds, relaxing on the order of 20 seconds. Relative spatial distributions of the 3 compression levels on a small torso area (i.e., right above the hip covering actuators 9-12 were measured with Tekscan CONFORMatTM (Figure 3-20). The region was selected since that area is relatively flat on the body, the reduced body contour allowed for better pressure sensor positioning. Note that no scale is provided: the measurements presented only provide relative comparisons and snapshots of pressure topography/distribution. This is due to difficulties in calibrating the system to provide absolute pressure measurements which continue to confound the ability to accurately measure absolute pressures on a garment-body interface. This is especially so for highly contoured body areas; most suggested calibration methods are limited to limbs where an air bladder can be fitted over as a 69 reference measure (e.g., arms, legs) [169], [170]—these developed calibration methods are currently not possible for areas like the torso and shoulders.

A B

C D

Technology Development Process- Summary

To summarize, this chapter presents the end-to-end cycle of the design and development process for wearable compression garments integrated with soft robotic actuation technologies (i.e., SMA actuators in coiled configurations). Two pilot user studies were presented as part of the iterative design process to understand the design space for dynamic compression technologies that uses soft garment structures, integrated with SMA actuation technology. The first being a low fidelity prototype exploratory study, followed by a medium fidelity SMA integrated prototype feasibility study, and finally resulting in the design of the high fidelity SMA integrated garment prototype that can/will be used as a research tool to further study the experiential effects of compression as a haptic modality.

Since a user-centric, iterative design process was employed by gathering and integrating user feedback in the development process, we were also able to better understand and characterize the design space for SMA-driven compression garment technologies. The user studies allowed the synthesis of design/engineering lessons learned in relation to soft-robotic and garment structure coupling and established the set of parameters that could influence and define the user experience of compression. In addition, the studies also revealed other interesting questions for future investigation (e.g., the potential advantages of the presence of warmth in simulating human touch).

The SMA actuators produce both compression and warmth simultaneously, which may be better at replicating physical sensations and components of human touch (e.g., hugs) than compression alone. Further, we also encountered participants’ novelty effect of wearing the dynamic garment (i.e., we may have a group of participants who feel calmed by the garment and on the other end of the spectrum, wearing the active garment may produce feelings of excitement due to the novelty of the technology/sensations). All of these involve wide implications in the general field of wearable haptics and should be further researched.

Table 3-3 presents a comparison of the three garment iterations. The success in this new domain of technology innovation will not only enable new types of on-body experiences, but also allow the possibility of addressing research questions pertaining to the experiential effects of on-body compression stimulation. Specifically, the successful development of these compression garments bring us one step closer to our goal of studying the effects of varying compression inputs on the body (such as changing intensity, duration, locations, and pattern), and enables the design of novel applications that may not have been possible before.

Table 3-3: Low to high fidelity garment iterations [167]

MEDIUM FIDELITY HIGH FIDELITY LOW FIDELITY PROTOTYPE [164] PROTOTYPE [168] PROTOTYPE [137] Active SMA Compression Active SMA Compression Passive Controllable (Tethered / Manually (Untethered and Wirelessly Compression Controlled) Controlled)

Garment Image

Ways of integrating SMAs Towards realizing a fully Quickly gather user into a garment structure to contained, remotely controllable, Objective experience feedback on apply compression on the computer-mediated, dynamic compression stimuli. body. compression system. - Application of varying - Rapid prototyping compression location, intensity, - Adjustable hook-and- - Application of compression and duration loop straps on varying body locations System - Two-way compression-expansion - Low effort in - Utilizes active materials to Features - Remote-control capabilities prototype construction apply compression on the - Full system portability - Easily sourced body - Included male / female gender materials garments - Tethered to a power supply - Only provides one compression - Requires manual - Only allows a binary on/off pattern / rate and has to be compression compression program manually controlled to achieve a application and - Utilizes inflexible Teflon targeted compression pattern System adjustment fabric materials for heat - Still relatively heavy for everyday Limitations - Utilizes stiff canvas protection use (battery bulk) fabrics (low comfort) - Only had a one-way - Does not include a feedback - Poor fit and comfort compression system with integrated sensing - Design / fit men only - Design / fit men only capabilities

Chapter 4

4. Variables Influencing User Experiences of Computer-Mediated Compression

In Chapter 3, we established the design space of garment technologies that can apply compression on the body, specifically presenting the iterative development process of soft garments that use shape memory alloy (SMA) actuators to present distributed compression on large body areas. However, to optimally deploy compression as a haptic modality in future applications, the effects of compression on the human body must be better understood. In this chapter, we present an interaction-focused experiential user study that aims to understand user perception of various compression parameters, through the deployment of the developed high-fidelity SMA-based compression garment system (Figure 4-1), capable of delivering spatially and temporally dynamic compression. Specifically, the goal of this study is to assess if and/or how subjective experiences (e.g., preferences, emotional impacts, comfort perceptions) vary among individuals subjected to a variety of compression-based stimuli.

Figure 4-1: The high fidelity SMA-integrated garment worn by a female. SMA actuators with controllable intensity levels were located on the torso, arms, and shoulders regions.

Interaction Study Methodology

Given the development of the high fidelity prototype (Figure 4-1) as a result of the design process outlined in Chapter 3, the technology enables the study of the effects of changing compression parameters on users’ experiences. With that, a within-subjects user study was conducted (n=17, 8M/9F, aged 18-29, average age= 22.1 years old; participants were recruited from the university through flyers and word-of-mouth); four compression variables were systematically investigated:

1. Location- depending on where the SMA actuators are activated on the upper body 2. Intensity- contingent on how many actuators are activated at a given instance 3. Duration- given how long the actuators are powered 4. Pattern- constant compression when actuators are powered for the entire presentation duration or pulsing compression when actuators are sequentially turned ‘on’ and ‘off’

For location, compression was applied on three body regions: torso, shoulders and upper arms. For shoulders, from the low and medium fidelity studies presented in Chapter 3, the applied shoulder actuator orientations were reduced to 2 options: ‘shoulder-straight’ (SMAs oriented vertically) and ‘shoulder-diagonal’ (SMAs oriented diagonally towards the chest). Three levels of intensity (low, medium, high) were applied at each body location. The study was pseudo-randomized. The ‘torso’ condition with 3 intensity levels was performed first since shoulder test conditions were contingent upon anchoring of the trunk (otherwise the garment will ride up as shoulders SMAs are actuated). The participants selected a preferred torso intensity before shoulder compressions were applied; for 74 that, they received a randomized shoulder orientation group before moving to the next shoulder group. Armbands were added when torso and shoulder preferences were decided. Each body location included 3 compression intensities in random order, giving 12 test conditions (4 locations/ orientations x 3 intensities). In each condition, the participants donned the loose-fit, unactuated garment and then provided compression for 1.5 minutes (since pilot studies showed users were able to detect compression sensations within seconds and form a preference within a minute) [164].

After the 12 test conditions were done, compression timing, i.e., ‘duration’ was investigated to understand how a lengthier exposure affects a user’s compression experience. In this test, the garment—having each user’s preferred location and intensity settings dialed in—was actuated for a maximum of 10 minutes (this timing was selected since it is a common duration of compression therapy, e.g., in occupational therapy where compression is applied for extended time as a method of relaxation/focus (e.g., DTP) [4], [12]) by which participants were asked to voice if and/or when discomfort was felt. The ‘duration’ test ended when participants felt discomfort from the system or at the maximum duration of 10 minutes. Throughout the ‘duration’ test condition, participants viewed images of everyday objects (from the International Affective Picture System (IAPS) database [171]) as an emotionally neutral activity to occupy their attention, as well as to prevent users from ending the ‘duration’ conditions simply out of boredom. The final condition involves contrasting stimulus pattern as all previous conditions—instead of a constant compression, stimulus was ‘pulsed’ on/off every 30 seconds for 3.5 minutes (made possible with the re-expansion of braided sheath but was controlled manually, i.e., the Processing user interface buttons were turned on and off sequentially) and subject preferences were studied qualitatively.

The participants were seated during the test and completed surveys probing their perception of compression intensities, comfort/discomfort on various body locations, perceived emotions, and stimulus parameter preferences. Participants were also asked to ‘think aloud’ throughout the study to form a better understanding of their thought processes. We also collected physiological signals (electrodermal activity (EDA), heart rate, temperature) with an Empatica E4 wrist device; note that this was not a main goal of this study, it was only done to record any incidental observations in response to the compression stimulation and investigate if there were potential objective effects of compression on the body. At the end of the study, qualitative interviews were conducted to better understand each user’s experience of compression.

Interaction Study Results and Discussion

The survey results, supplemented with think aloud remarks by participants, are presented (n=# denotes the number of subjects that made similar comments). Qualitative interview results gathered were coded, organized thematically, and presented in the following section categorically.

Perceived Compression Intensities

The perceived compression intensities on various body locations (Figure 4-2) were collected on a scale of 0 (no pressure) to 10 (max. intensity); subjects were instructed to consider 5 (moderate) as equivalent to an average hug. The perceived compression intensity rating results are presented in Figure 4-3 to Figure 4-5. To determine if there were significant differences between perceived intensities of low, medium, high, for each body location, non-parametric Kruskal-Wallis rank sum test was employed, and significant differences observed are indicated in the associated figures. The full summary statistics can be found in Appendix A.

As shown in Figure 4-3, there were no quantitative increases in perceived intensity for the torso regions corresponding to the actual applied compression inputs of low, medium, and high. In the lower back, especially, high-intensity compression resulted in lower compression perceptions. Subject comments were nonetheless distinguishable by some among the levels; low torso compression was described as subtle (n=3) and high torso compression was associated with more compression/restrictive sensations (n=6)). Three factors may shed light on this paradox: (i) Many (n=13) commented that the compression mostly started and centered on the abdomen and sides (especially since soft tissues undergo larger volume changes with breathing/movement), possibly drawing attention away from the lower back. (ii) Since the scale was anchored at 5-moderate intensity as being similar to an average hug, the magnitudes of the scale may not be sufficient to

76 tease out the subtle differences between conditions when compared to ‘an average hug’ (which could be highly variable—also evidenced by the high standard deviations seen in Figure 4-3). (iii) Also, garment sizing and fit were the most challenging in the torso region: The garment did not fit well on 5 out of 17 subjects (as determined by participant self-report and researchers’ observations); two females had a snug fit in the torso, one male wore the female-sized garment due to his small stature, and the other two females wore male-sized garment because their torso/chest did not otherwise fit. The sizing and fit of the garment likely played a role in the distribution of the intensity perception results; with overly snug or loose initial fit, applied compression differentials will be less evident.

Figure 4-3: Perceived compression intensity ratings for Torso (sides, abdomen, lower back) and Arms. The lower and upper box boundaries are of the 25th and 75th percentiles; the center line represents the median; the center cross marker indicates the mean; the whiskers are of the 10th and 90th percentiles. (Significance codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’ , 0.1 ‘.’)

In contrast, the arms showed a different picture; the Kruskal-Wallis test showed potential differences between intensities, χ2(2) = 5.295, p= 0.071. While not strictly ‘statistically significant’ under typical definitions (likely due to the large variability in our small sample), we still report the statistical results based on an assumed threshold for probability of type 1 error < 0.1, since many participants reported experiencing differences between intensity levels. Post hoc analysis using pairwise comparisons using Wilcoxon rank sum test with p-value adjustment using the Benjamini- Hochberg (BH) method (to control for false discovery rate) showed potential differences between low and high intensities for the arms (p=0.082). This aligns with participants’ think aloud 77 comments, 9 subjects mentioned that low level arm compression was felt but subtle, and there was an increasing number of comparisons to a blood pressure (BP) cuff with increasing intensity (nlow=1, nmed=3, nhigh=7).

Figure 4-4 and Figure 4-5 present the perceived compression intensities on the upper chest, back, and shoulders, given ‘straight’ and ‘diagonal’ shoulder compression vectors. Generally, the perceived intensities were more distinguished compared to the torso, especially on the top of shoulders (both compression vectors); for the high intensity, subjects (nstraight=7, ndiagonal=7) voiced clear distinction that it gave the most compression. Further, majority of the subjects were able to distinguish between compression vectors. With ‘straight’ vector conditions, participants felt more compression on the upper back/shoulder blades (nlow=3, nmed=3, nhigh=8) while ‘diagonal’ vector was said to provide compression starting and remaining more on the front shoulders and/or chest

(nlow=7, nmed=6, nhigh=7). This is reflected in the rating results, Kruskal-Wallis test showed distinctions between compression intensities for ‘straight-back of neck’, χ2(2) = 5.791, p= 0.056 and ‘diagonal-top of shoulders’, χ2(2) = 6.939, p= 0.031. Post hoc analysis using pairwise comparisons using Wilcoxon rank sum test with BH p-value adjustment showed potential differences in perceived ratings for ‘straight-back of neck’ between low-high (p=0.072) and med- high intensities (p=0.072) and between low-high for ‘diagonal-top of shoulders’ (p=0.026), as indicated in Figure 4-4 and Figure 4-5.

However, we also note some upper body areas where no compression intensity differences were observed (e.g., front side chest, front mid chest, upper back). Those body regions are likely less affected (with less differentiable compression sensations) due to actuator positioning; the SMAs were situated on the shoulders, hence the compression sensations were likely not translated to the listed body regions that are further away). Overall, there were observed differences between low and high intensities for the arms, top of shoulders, and back of neck (i.e., directly correlating to SMA integration locations, hence compression is most felt). However, even in those locations, the medium compression intensity is sometimes not well distinguished—which could mean that the perceived differences between intensity levels may be below some participants’ discrimination thresholds. If all three distinguished intensity levels are necessary for future applications, psychophysical studies have to be performed to more systematically quantify those differences and incorporate relevant strategies based on design needs (e.g., modifying the number of parallel actuators activated at an instance for each compression level).

Compression Comfort/Discomfort

The comfort/discomfort of varying compression stimuli on body locations was investigated using the Comfort Affective Labeled Magnitude (CALM) scale [172]. The CALM scale contains a ±100 labeled magnitude scale of comfort/discomfort; the adjectives ‘greatest imaginable’ defines the fixed end points of positive and negative comfort experience. The scale was selected since it was designed to be sensitive to a wide-range of comfort-related stimuli including garment types with demonstrated good reliability, but perhaps more importantly, since participants select an arbitrary number along the scale, it is possible to treat this as ratio data [172]. Since there are major anatomical differences between males and females, Figure 4-6 and Figure 4-7 display gender- separated means (triangle marker: males; circle marker: females) and standard deviation (with unidirectional error bars to prevent visual clutter) to capture possible gender-based comfort perception differences. ‘Baseline’ comfort was collected prior to wearing the garment (no applied compression); ‘duration’ condition had subjects exposed to their most preferred stimulus features for a maximum of 10 mins. One-way analysis of variance (ANOVA) was performed to determine if there were significant differences between conditions for each body location; due to the small sample size, this statistical analysis was not gender-separated; significant differences observed are indicated in the figures below. The full summary statistics tables can be found in Appendix B. Although not all body regions showed significant differences in comfort ratings between different compression intensities (likely due to a combination of factors including small sample size and subjectivity/qualitative nature of this study), we still highlight and discuss the general trends observed because it is important to understand how comfort changes with different compression intensities and body locations that are more/less impacted by the application of compression.

Figure 4-6 shows the compression comfort of various torso regions and arms. Broadly, we see a trend of reduction between baseline (blue) and compression conditions, but they are still positively rated. In other words, the compression application generally is less comfortable than the garment unactuated, but still within comfortable ranges. However, we noticed an exception to this inverse intensity-comfort trend: in the middle/lower back, compression did not drastically reduce comfort levels compared to baseline. In fact, while all other torso/arm areas have an inverse intensity- comfort relationship, lower back area was otherwise (for males, highest compression intensity resulted in lower back comfort ratings to be higher than baseline). This could mean two things: (i) subjects may be more inclined towards higher compression on the lower back; many subjects

80 verbalized preference for lower back compression since it feels supportive/aids posture (n=5), or (ii) this reflects the aforementioned reduced perception of lower back compression corresponding to higher system-generated compression intensity (n=13 indicated compression on the torso were mostly felt on the abdomen and sides, which could have drawn attention away from the lower back). In light of this, a more ‘balanced’ compression design should consider moving the actuators slightly towards the back so that the lower back compression can be more evident than the front.

For the arms, the ANOVA analysis showed significant differences between the group conditions, F (4, 72) =6.041, p <0.01. Post hoc analyses with the Tukey multiple comparison test revealed comfort rating differences between baseline/duration and medium intensity/high intensity; the results are visually indicated in Figure 4-6 and presented as a summary table in Appendix B. Broadly, we again see that as intensity increases, comfort ratings decrease; reminders of blood pressure cuff (n=11) with increasing compression likely influenced participants’ comfort ratings. However, this effect was not seen in the ‘duration’ condition as participants were allowed to select their most preferred arm compression intensity (this includes ‘no compression’ conditions).

Figure 4-7 presents subjective comfort on various (higher) upper torso locations and shoulders, the ANOVA results showed significant differences between conditions for the top of shoulders, F (7, 127) = 2.237, p= 0.0353. Post hoc analyses with the Tukey multiple comparison test revealed comfort rating differences between high intensity and baseline as well as duration; the results are visually indicated in Figure 4-7 and a summary table is in Appendix B. What is most interesting is that we observed possible gender differences in perceived comfort. While the female subjects’ comfort ratings did not reveal large differences on applied compression intensity for different vectors, male subjects generally found high upper torso/shoulder compressions less comfortable; when compression on those areas were less intense, it was said to ‘feel calming/nice without grabbing too much attention’ (n=1), but with increased intensity, breathing/moving were said to be much more restricted (n=4). Males were able to tolerate compression for the ‘straight’ vector at higher intensities compared to ‘diagonal’ (participants generally tolerated ‘straight’ conditions at medium levels and below, but ‘diagonal’ was reported as significantly less comfortable at all settings greater than low intensity). With ‘straight’ vector, participants described feeling compression mainly on the upper back/shoulder blades (n=14), whereas ‘diagonal’ vector was mostly on the chest, consistent with the comments of feeling restricted.

For all body locations, comfort ratings for ‘duration’ (gray) most closely resemble the ratings of baseline, likely because users were asked to select preferred compression settings. The increased comfort when preferences are accommodated (i.e., baseline comfort was most closely mimicked when users were allowed to self-select preferences) highlights the need for system customizability in optimizing subjective comfort.

Stimulus Preferences/Liking

We also asked participants to rate liking/disliking of compression for each condition (Figure 4-8: green-liking; blue-neutral, red-dislike), as well as their most preferred stimulus parameters (Figure 4-9a). Figure 4-9b shows torso intensity preferences, broken down by gender and personal preference for hugs (probed during baseline). Liking/Disliking of Provided Compression

Duration 1 4 7 4 T + Sh. + Arm High 1 1 3 6 3 3 T + Sh. + Arm Med 2 1 3 3 6 2 Dislike a great deal T + Sh. + Arm Low 1 2 3 9 2 T + Should. Diagonal High 3 3 5 4 2 Dislike a moderate amount T + Should. Diagonal Med 3 2 8 4 Dislike a little bit T + Should. Diagonal Low 2 3 4 6 2 Neither like or dislike T + Should. Straight High 1 3 4 5 3 1 Like a little Test Conditions Test T + Should. Straight Med 2 4 5 4 2 T + Should. Straight Low 3 3 3 7 1 Like a moderate amount Torso (T) High 3 4 3 6 1 Like a great deal Torso (T) Med 1 5 4 7 Torso (T) Low 1 3 6 7

0% 20% 40% 60% 80% 100% Percentage of Participants Figure 4-8: Liking/Disliking of garment compression rated on a 7-point scale. The frequency of participants for each category are indicated in the figures.

From Figure 4-8, we see a large percentage (80-90%; indicated as blue-grey to green in the figure) of our participants were not averse to any of the compression presentations. The dislikes were mostly consistent with the comfort ratings whereby higher intensities drew more negative subjective stimulus perceptions. One thing to note is that none of the participants disliked compression presentations for the longer duration study, likely because they were allowed to customize their preferred compression inputs.

From Figure 4-9a, for torso regions, our subjects had stronger preferences for either the low or high intensity compressions. When broken down by gender and personal preferences for hugs (Figure 4-9b), males tended to gravitate to the high intensities for torso; yet for shoulders, almost all preferred the low intensity (n=2 no shoulders), regardless of compression vector. In contrast, for females, when ‘straight’ vector was applied, high intensities were preferred, and when ‘diagonal’ vector compression was applied, the decision was split between low and high intensity. In general, the medium intensity on the shoulders was the less popular choice. As for arms, participants generally liked either a low/medium compression, or none.

Preference Frequency Test Condition Males Females Total Low 6 Torso Medium 4 High 7 Low 3 Baseline Torso Preference Intensity Male Female Straight Medium 2 Low 1 4 Like hugs Shoulders High 3 Med 1 2 Low 5 ☺ High 3 3 Low 1 - Diagonal Medium 0 Don't like Med 1 - High 2 hugs  High 1 - No shoulders N/A 3 Low 3 Arms Medium 4 High 2 No Arms N/A 7

Figure 4-9: (Left (a)) Compression stimulus presentation preference; (Right (b)) Torso compression stimulus intensity preference by gender and personal preference for hugs

We know that the perception of the compression garment is also driven by thermal comfort. Since the actuators are activated through Joule heating, thermal comfort is an important facet of this garment, even though thermal stimulus was not the main focus of this investigation. The results of temperature liking/disliking are presented in Figure 4-10. Unsurprisingly, the thermal comfort of the garment was impacted since the SMA actuators generate heat; together with the multi-layer construction of the garment, it likely drew more negativity than a regular garment would. From Figure 4-10, the most compelling takeaway is that when we compare it to Figure 4-8 (compression), thermal comfort is drastically reduced in the ‘duration’ condition since heat accumulates over time.

Figure 4-10:Liking/Disliking of garment temperature rated on a 7-point scale. The frequency of participants for each category are indicated in the figures.

Subjective Comfort/Discomfort Thresholds

For ‘duration’ condition, since we probed participants every minute and encouraged comments, we broadly determined compression and temperature subjective discomfort thresholds (thermal comfort was also considered since SMAs generate heat). For compression, 11 participants voiced the tipping point into slight discomfort at around 8.5 minutes, yet it still ‘felt good’ and the sensation was compared to weighted blankets or hugs (n=3). All participants noted the applied compression was relatively constant throughout the period, and 3 said it was mainly noticeable during breathing but less so otherwise. In contrast, the averaged discomfort threshold for temperature was 4:56 ± 0.08 minutes; only 3 subjects enjoyed the heat and significantly more complaints were voiced since heat accumulates over time. While compression remained relatively constant, all participants mentioned they could feel the temperature rising throughout the period, likely due to a combination of the heat generated, as well as poor breathability due to the multi-layer garment construction.

Pattern of Compression Stimulus

In the final condition, we investigated how different temporal patterns of compression (‘constant’ vs. ‘pulsing’ of 30-seconds on-off) affected users’ experiences. Results on preference were close to being evenly split with 8 participants preferring the constant compression, 6 preferred ‘pulsing’. Additionally, n=3 stated that the preference was situation dependent. Collectively, those who preferred the constant compression noted that it felt secure (‘like wrapped in a blanket’) and was 85 less attention-demanding (‘[you] forget about it when it’s constant’), yet such constant compression was inviting only with a gradual stimulus ramp-up, since ‘sudden change is off-putting’. In contrast, those who preferred ‘pulsing’ stated that they liked the renewing sensation and thought that it was more relaxing and akin to a shoulder massage (‘…like a non-violent massage chair’; ‘…would help me fall asleep’). It was also described as being less intense (i.e., more room for breathing). The 3 subjects who cited situational dependence indicated that the pulsing would be used for stress relief while the constant compression would help one feel ‘ready/prepared’. These differing connotations (i.e., constant stimulus provided feelings of active/security; pulsing was associated with relaxation), have interesting implications in future designs since compression patterns could provide contrasting perception/emotions, and we can envision the breadth of functionality they provide (e.g., eliciting diverse sensations in immersive environments: receiving a hug for support vs. muscle enhancement in gaming).

Perceived Emotions and Subjective Emotional Changes

Since it is known that varying emotional experiences could arise given different forms of touch, we were also interested in participants’ perceived emotions resulting from this form of SMA-actuated compression application. The Self-Assessment Manikin (SAM) was employed [173]; it includes three 9-point scale questions (with human-like figure accompaniment at each point), based off the vector model of emotion (valence, arousal, dominance). The valence scale ranges from positive to negative, arousal from excited to calm, and dominance from dominated to dominant (for a numerical scale of 1-9) [173]. From Figure 4-11, for valence, participants’ emotions were skewed towards positive valence, with the median being either 3 or 4 for all test conditions. For arousal as in Figure 4-12, the median was either on 5 or 6 for all conditions; but the range was the largest compared to other measures of emotion. While the general trends were towards neutral-calm, we note that some participants could perceive this stimuli as being ‘exciting’. For dominance (Figure 4-13), the median was 5, with neutral dominance measures for all conditions.

From participants’ qualitative interview, a majority of the participants felt that the first garment compression experience was exciting/strange/silly/weird (due to its novelty and being a ‘foreign experience’), but they soon adapted and was more comfortable as the study progressed (n=10). Overall, the majority of the participants (n=11) reported that the garment made them feel more comfortable/relaxed/at ease (consistent with general trends in SAM survey), 2 participants explicitly voiced they felt more positive/happier, 1 felt more active (especially with sustained arm compressions that increased dominance). While 5 participants stated that they were far more comfortable after finding the right personal settings with the garment, 3 participants stated that since the study was relatively long, the stimulus felt uncomfortable/unnatural midway, and the garment was not something they would wear daily. This particularly applied to when the garment felt too tight/hot, they in turn felt anxious/alarmed (n=2). When asked to describe their mood after garment removal, 6 participants described the sensation as feeling lighter, more free (‘was too hot and heavy’), and relieved after it, while others felt as though they have more energy (‘feel refreshed’, ‘had a boost’). Two of the participants who preferred the heat (self-describing as naturally ‘cold’ persons) voiced discomfort due to lack of warmth after garment removal (‘felt good to have it on’).

Comparison to Physical Touch

When asked to describe how the experience of the garment compares to other forms of touch (e.g., hug), 6 participants stated that the sensations included areas not typically felt in hugs (e.g., arms/lower back) but have the same physical elements. 3 subjects felt the compression was more consistent than a hug since it was all around (‘feels much more ‘whole’ than a hug’) and feels nicer in instances where compression was comfortable. The garment was also said to be closer in proximity than hugs since hugs may have some physical gaps. However, some mentioned that the sensations provided by the garment was not as good as a hug psychologically, because it lacks emotional factors (‘feels impersonal’) and felt less natural due to it being an object (n=4). We observe a wide spectrum of opinions on the perception of a computer-mediated compression, likely because no context was presented alongside the stimulus.

Word and Scenario Association

To understand how the generated compression sensations might/could be interpreted, participants were asked to use several terms to describe their feelings during compression as well as situations they might expect the garment to be used. We identified 3 common word categories (Figure 4-14):

(i) calming/relaxing/comforting, (ii) restricted/tight/secure, and (iii) warm. The dichotomy of ‘calming’ vs. ‘restricted’ is interesting, and these results align with the scenario associations (Table 4-1). Most participants associated garment usage with emotion-related situations (16 instances), such as stress/anxiety relief, since the garment felt calming or hug-like—relating to secure/tight connotations. The 2nd most prevalent theorized application is medical use (10 instances) (e.g., physical therapy), where the ‘secure/restricted’ fits. These results further suggest that compression stimulus will be received and interpreted differently depending on its use context and expectations.

Figure 4-14: Free Word Association Word Cloud .

Table 4-1: Scenario association in which situations participants expect the garment to be used are summarized and presented

Category Function Used during anxiety/stress relief/calming (n=11); Emotional therapy (n=3); To feel Emotional (n=16) safe/simulate hugs (n=2) Back support/Brace (n=3); Autism/sensory therapy (n=2); Pain relief (n=2); Medical (n=10) Physical therapy (n=1); To encourage circulation (n=1); Used during menstrual cramps (n=1) Physical-Pressure (n=4) Massage/muscle relaxant (n=3); Athletic garment (n=1); Military use (n=1) Physical-Heat (n=6) Muscle heating (n=2); Body warming (n=4)

Overall Garment Comfort, Preferences, and Rating

Overall, participants provided their experience with the garment an averaged rating of 6.76 ± 1.03 (out of 10). In general, participants enjoyed the compression provided, with some stating the garment “grew on them” (n=4); once all parameters were set to their liking, the garment was said to be more comfortable than it being unactuated. In terms of general design of the garment, some 89 participants commented on the garment’s material stiffness (n=3); woven canvas was used in the inner layer to provide structure and efficient compression but was stiffer compared to everyday garments. Some (n=3) hence stated that its snugness could be either useful or restrictive depending on the activity being performed in (freedom of movement is compromised). Therefore, a clear design insight is the relationship between functionality/context and comfort; for instance, a structured garment whose function is to restrict movement (e.g., brace) is likely to have a shifted comfort perception than a structured garment whose stiffness shows no clear function.

In terms of compression distribution, participants (n=6) had most complaints about the chest area; it was said to be uncomfortable since deep breaths were constrained. Many requested to have more compression on the back (nupperback=5, nlowerback=5) provided it be ‘supportive and not overly tight’. As for the arms, many participants felt the sensation was “weird” (n=7) since it reminded them of blood pressure cuffs, yet others felt the arms provided different/enjoyable sensations (n=3), much so that 2 subjects wanted arm compressions down the full length of their arms as long as mobility is not restricted. In terms of compression intensity, the consensus was that weakest compressions were subtle, and the strongest could easily be felt but wasn’t overbearing. However, participants still had unique preferences for the way and how much compression was applied, as one subject pointed out: ‘the one I picked was just right, feels like a group hug’; implying a need for customizability of stimuli parameters. In terms of thermal comfort, many commented it ‘was good for a short time, then it starts to creep up [and be annoying]’ (n=3 indicated the heat was not unbearable; n=4 said it was too hot). However, 3 subjects enjoyed the heat provided by the garment, describing themselves as naturally ‘cold’ persons. Interestingly, heat in lower back was again mentioned most frequently throughout the study (29 unique instances); this is intriguing because none of the SMAs were located at that location. We likely observe this since the back/spine is said to be a high heat/sweat zone [166]. While we anticipated the need for ventilation and used foam- mesh material and vent channels on the back and insulated the SMAs with the braid mesh material, the issue of heat management remains evident.

Physiological Changes Brief Findings

We also obtained preliminary insights from the psycho-physiological changes associated with compression stimuli using an Empatica E4 device (Figure 4-15). Note that as explained, physiological changes were not a major focus of this study and was only investigated to note incidentally interesting observations (towards a closed-loop system to monitor one’s physiology to 90 inform the application of compression parameters) and/or drastic changes physiologically that could be a cause of concern. The major observation noted was that the physiological data of participants were widely different across individuals. Here, we specifically draw attention to a qualitative case study with 2 male participants and shed light on the implications of the observed variations between participants.

Two participants (P13, P15) were selected because they had a great initial garment fit but differing self-reported preferences for hugs (this is important since an inclination for hugs or lack thereof could result in differing perceptions of on-body compression due to the mechanical similarities in sensations). Specifically, we pay attention to changes to these partipants’ electrodermal activity (EDA) as a preliminary investigation of emotional arousal. The colored circles on the x axis of Figure 4-15 denotes compression intensity; the hearts on the top denote each participant’s favorite compression parameters; the red vertical lines represent the start of every test condition.

Participant #15 - Self-identified as one who “Likes hugs” ☺

Baseline Torso T+ Shoulder Diagonal T+ Shoulder Straight T + S+ Arms Duration

Participant #13 - Self-identified as one who “Doesn’t like hugs” 

Baseline Torso T+ Shoulder Straight T+ Shoulder Diagonal T + S+ Arms Duration Figure 4-15: Physiological data case-study with two participants; P15 and P13. The red vertical lines dividing the physiological data divides the different test conditions experienced by participants; the heart marker indicates each participant’s self-reported favorite compresison parameter. Note that P15 did not indicate a preference for arm compressions.

From Figure 4-15, we note that P15’s EDA displays large phasic skin conductance response (SCR) immediately following compression stimulus and the spikes gradually decay, especially in conditions that he prefers (P15’s SCRs were drastically reduced in the latter conditions, especially arms, which he disliked). In contrast, P13, while also having distinct preferences regarding the varying stimuli parameters, did not show similar drastic changes in SCR activity throughout the duration of the study. This demonstrates the case of individual differences in EDA activity (some fraction of the population is likely to have slower electrodermal reactivity, whose SCRs rise very 91 slowly), making the process of finding a generalizable detection system for something subjective such as preference/comfort particularly challenging. Further, while EDA reflects arousal, however, this method does not provide information on the valence of the stimulus (i.e., how positive-negative one felt). Therefore, objective biometrics such as EDA data will require corroboration and/or contextualization in the form of other objective biosignals or subjective self-reports in order to be useful in specific applications. One example application could involve the detection of a negative, high arousal emotion, in turn triggering compression stimulus delivery to calm a user down (note: a user who is receptive to this form of stimulation)—a concept similar to those used in DTP therapy.

Interaction Study Discussion and Implications

This study investigated the effects of various parameters on the experience of compression with an SMA-based dynamic compression garment. We discuss broader themes and implications for future garment design or applications, and synthesize lessons learned about how potential confounding stimuli arising from the design of the garment actuation strategies may influence user experiences.

The Influence of Compression Parameters on User Experience

The first major theme gathered is that varying compression parameters had an influence on the physical experiences of users. Participants were able to clearly distinguish between ‘straight’ and ‘horizontal’ shoulder vectors with varying intensities. For instance, shoulder ‘straight’ was said to have generated more downwards compression sensations than ‘diagonal’ (which felt more medial towards chest). The language used to describe these sensations were also varied; participants (n=4) described ‘straight’ vector as ‘someone squeezing/placing hands on the shoulders’, but ‘diagonal’ was described as being ‘less mobile/more restrictive on the chest’(n=5). Applications that use compression require an appreciation of the part these variables play; depending on the activity being performed, the snugness could either be beneficial (e.g., provides structured back support like a brace) or restrictive (e.g., limits range of motion). And the use context should also be considered for optimal effectiveness, for example, a stress-relieving compression system will likely require a design that reflects calmer/subtler compression; having compression vector on the chest where breathing turns restrictive may be counterproductive. The relationship between functionality /context and comfort is important; understanding the role these parameters play will help improve future system designs.

We also noticed the perceived compression intensity results were much clearer on the arms and shoulders than on the torso region. In fact, when high-intensity compression is applied, lower back resulted in lower compression perceptions. We think that this most likely because a large majority of participants (n=13) commented that compression was mostly on the sides and abdomen (where soft tissues undergo larger volume changes with movement), possibly drawing attention away from the lower back. This finding is consistent with past studies, such as the one by Delazio et al. who found that the shoulders were the most sensitive to compression relative to other areas such as various torso regions including the chest and back [17]. Therefore, future designs should account for these relative sensitivities’ differences corresponding to the various body regions if the goal is to deliver significantly distinct compression stimulus intensities.

Gender Differences and Sizing/Fit

However, the differences in perception noted given compression stimulus on varying body regions may also be modulated by a secondary variable—garment sizing and fit. The male and female anatomy differs; yet the role these dissimilarities play in the experience of stimuli is under- appreciated; most of the devices constructed and used in the literature does not account for the role gendered anatomical variances bring, possibly as confounding stimuli. Even when separate garments were made for each gender to account for anatomical variances and limited to only one size (size M), we still had sizing/fit issues. The garment did not fit well on 5 out of 17 participants (based on a combination of researcher observations and participant self-report), with female fit issues on the chest area and male issues on the length/area under arms (albeit to a lesser degree). In terms of physical anatomical differences, the chest region likely brings the most significant impact, especially for designing these conformal garments. And since we are designing for the upper body, the fit problem is over-represented and amplified in this study.

When fit is compromised, the garment may shift, presenting stimulus on altered body areas or producing diminished response due to actuators failing to contact the body (if too loose); or reducing functionality (initial fit that is too tight will reduce sensitivity to additional actuation; as described by the Weber-Fechner Law [55]). We can also explain this from the lens of the known mechanics of the soft robotic system (as defined in Chapter 3.2.2)—compression is generated because of an antagonism between SMA actuators and the fabric, which creates an equilibrium tension. This equilibrium is established based on the relationship between the garment and the body, therefore, having the same garment on different bodies could give rise to different equilibriums 93 unless design strategies are implemented to account for these variances. Some design strategies could include the use of alternate materials and/or implemented methods of adjustability. We cannot expect a universally sized system to precisely administer on-body stimuli across a population, both physically and experientially (especially when there’s evidence with this study suggesting there might be gender differences in compression experiences and preferences). We observed gender-separated trends in compression parameter preferences. In general, we found that males tended to gravitate towards higher torso intensity yet lower shoulder intensity (regardless of its vector); for females, low torso intensities were favored, and the preferred shoulder intensities were more reliant on the applied vector.

Individual Preferences and the Need for Customizability

Compression parameters are also tightly intertwined with preferences and comfort perceptions— both of which are key factors for long-term wearability and acceptance. While general design insights can be drawn, a recurring theme was that user preferences vary across a spectrum and satisfaction/comfort is largely dependent on user’s ability to customize settings (e.g., Figure 4-6, baseline comfort was closely mirrored when users were allowed customization). An ultimate favorite setting across users did not emerge; however, preferences do exist, and were highly individualistic (i.e., they were unique to each user), including a subset of users whose preferences are universally negative (i.e., there are no compression profiles/distributions/ magnitudes that can be identified as enjoyable or positive). Once a user discovers their preferences, they generally responded more positively to compression stimulation (except those who universally dislike compression as a stimulus). Hence future systems should account for some level of customization and/or calibration based on its intended use. This means that different preferences and experiences of compression—both between genders and across individuals within a gender group—have to be accounted for and individually tailored to. For the time being, preference-finding is likely a participatory and iterative process, but a long-term goal would be to investigate methods for closed-loop (automatic) calibration using either physiological data (e.g., a combination of biometrics) and/or user feedback (e.g., experiential sampling through voice assistance) for an ultimate experience. All these together point to the need for users to be calibrated properly; we cannot assume that a universal system with a single set of stimulation parameters will be sufficient given people’s varied objective and subjective reactivities.

Relationships between Context-specific Stimulation and Perception

Moreover, we discovered that different stimulation parameters may invoke diverse emotions and perceptions. Broadly, participants viewed the garment as a potential strategy for emotion regulation or medical use and related it to contrasting terminology groups: calming/relaxing vs. restrictive/ tight. Such groupings suggest a range of possible effects given varied compression parameters and situational differences may call for varying stimuli on different body locations. To tie it all together, we investigate the following example: 3 participants stated their preferences were driven by the need for evenly distributed torso and shoulder intensities, but the arms were evaluated separately (likely due to the separation between trunk of body and arms by the shoulder joint). When these locations are investigated separately, the torso and shoulders have calming/relaxing connotations, but the arm compression was described as similar to blood pressure cuff or evoked feelings of activeness. Further, when we looked at varied compression patterns: constant and pulsing, constant stimulus was said to provide feelings of active/security, but pulsing was associated with relaxation. This has interesting implications in future designs since compression patterns on varied locations could provide contrasting perception/emotions, and we can envision the breadth of functionality they provide (e.g., eliciting diverse sensations in immersive environments: receiving a hug vs. muscle enhancement in gaming). The mapping between language and haptic space was performed with a pneumatic compression jacket by Delazio et al. [17], but have yet to be extended to active material (SMA) actuation schemes, and should be further investigated.

4.3.1. Limitations and Future Work

While the compression garment system achieved most of the specified design targets specified in Section 3.5, there are still areas that warrant further investigation. The first is the use of the braided sheath for cyclic ‘resetting’ of the SMA actuators. While we’ve witnessed considerable success in this strategy, one thing to note is the relaxation rate is dependent on braid stiffness, but it is in turn also mediated by the actuator force output and efficiency. In other words, a stiffer braid will allow a quicker relaxation rate, but will require more SMA force to contract, in turn using more power. To fully exploit the use of this strategy, a full characterization of the braid-SMA system should be investigated. At the current power ratings, the actuators contract on the order of 2-8 seconds and fully relaxing at about 20 seconds; this demonstrates an uneven contraction-relaxation rate. An optimal rate will of course depend on the application itself but if the goal is to achieve an equal compression ramp up (contraction) and ramp down (relaxation) cycle, various design strategies can 95 be considered including reducing the power supplied to the SMA (to slow the contraction rate), changing the braid stiffness including diameter, material, braid angle etc. (to increase the relaxation rate), or inducing a slight compression within the braid-SMA system by storing some elastic energy within the system to begin with (to increase the relaxation rate).

Secondly, while equipped with on-board power systems, the current prototype requires high power (battery) inputs which increases garment weight. Future work to optimize the system could involve fine-tuning material properties of SMA actuators to lower the thermal/current activation threshold (recent pilot showed promise in this avenue [174]), which will aid garment thermal regulation and comfort. Along those lines, thermal management remains a problem. Many participants commented that as the garment use duration was extended, the temperature became overbearing (three subjects enjoyed the heat, describing themselves as ‘cold’ individuals). A recurring theme still present in the high-fidelity garments was the heat experienced by participants in the lower back, even though no SMAs were located there. Alternative design considerations, including the use of different insulating materials, power modulation strategies (e.g., pulse width modulation (PWM)), mechanical latching mechanisms (to ‘lock’ the compression, enabling power to the SMAs to be shut off—but this will only allow constant compression sensations unless an ‘unlocking’ mechanism is introduced), and alternative placement of SMAs should be investigated to mitigate the experience of lower-back heat if intended to use for long-term applications. Further, in the current configuration, this prototype is capable of applying discrete levels of compression intensities. However, they are limited to a binary on/off program and does not include closed-loop feedback; incorporating several stimuli pattern profiles with pressure/physiological sensors that monitor and capture data will likely allow more flexibility with system use with enabled fine-tuned compression parameters. Finally, the option of a phone application will also serve to enhance the portability and use-experience of the system.

Interaction Study Summary

In summary, we designed a soft robotic wearable system capable of producing dynamic compression on the body. Major characteristics of the system include:

• Functionally capable of producing dynamic compression on the body • Allows varying compression inputs of location, duration, intensity, and pattern • Equipped with remote-control capabilities and full-system portability with on-board power 96

and electronics • Soft and unobtrusive form factor for enhanced wearability

Using this system as a research tool to understand the effects of upper-body compression, we summarize the following major themes and implications for future compression garment designs:

1. Gender differences and sizing/fit. We should not expect a universally sized system to present on-body stimuli equally (both physically and experientially) across a population, especially with the results suggesting there might be gender differences in compression experience and preferences.

2. Compression parameters influenced users’ experiences. For instance, shoulder ‘straight’ generated more user-perceived downwards compression sensations than ‘diagonal’ (medial towards chest).

3. Individual preferences and need for customizability. While there is no observed universal preference across a population, we observed self-report optimal preferences for most (but not all) individual users. Satisfaction is largely dependent on an ability to customize settings; hence some level of customization is necessary for creating experiences that are well-received,

4. The relationship between context-specific stimulation and perception. Broadly, participants viewed the garment as a potential strategy for emotion regulation or medical use and related it to contrasting terminology groups: calming/relaxing vs. tight/restricted. Further, constant stimulus was related to activeness/security, but pulsing had relaxation connotations. Such groupings suggest a range of possible effects given varied compression parameters and situational differences may call for varying stimuli on different body locations.

Other than as a research tool to further understand the effects of upper-body compression on a user, the developed technologies could also enable new modes of interaction between users separated by distance (e.g., tele-rehabilitation, social mediated touch) as well as new sensations in the area of immersive (AR/VR) experiences.

Chapter 5

5. Compressive Haptics in Emotional Communication

Touch plays an integral role in our everyday experiences [6]. Other than sensory-discriminative functions, the human sense of touch also serves an affective purpose, playing a part in how we interact with others [1], [3]. Interpersonal social touch is important from our early developmental years [7]–[9], and through the rest of our lives by reinforcing familial bonding, fostering romantic/ sexual relationships, all the way to influencing community social interactions [12]. In other words, touch provides us with a mode of communication, whether to convey emotions [5], [68], induce emotional experiences [11], or intensify other communication forms [15]. Given the myriad uses of touch, much effort is dedicated to developing artificial haptic devices to produce and study the salient features and nuances of touch [16]. More recently, the field of affective haptics is emerging as an intersection between emotions and haptics, encompassing technology that can detect, display, or communicate affect through touch [15]. In this chapter, our focus is on investigating the use of compression-based haptic devices for emotional communication.

Given the advances of compression systems presented in Chapter 3, we now have the technological

98 means that can enable remote communication—through the delivery of compressive and thermal sensations provided by SMA actuation. As evidenced by the preliminary findings in Chapters 3-4, the coupled compression-warmth sensations (henceforth referred to as ‘warm touch’) inherent to SMA actuation could potentially offer advantages in influencing affect due to their similarities to human touch. From Chapter 4, we also know that individual preferences vary in terms of preferred on-body stimulation using these SMA-based compression garments, but the design space for sending/ receiving emotional communication using ‘warm touch’ as a modality has not been extensively mapped. Hence, we want to go beyond what is possible (with the technology) and what is preferred (user evaluation/calibration necessities) to what is useful for communication. Thus, the goal of this chapter is to capture the strategies and mental models users would employ in cases where the SMA-based compression garments (capable of delivering ‘warm touch’) were to be used as a communication modality, as well as develop an understanding of user perceptions and behavior in order to design communication devices in this affective haptics space. This extends beyond just what/how ‘warm touch’ haptic parameters are used during emotional communication, but also include the whys, (i.e., what are the mental models used by participants during the ‘warm touch’ communication process). In order to achieve the goal of parsing different communication frameworks and their associations with emotion communication in relation to ‘warm touch’, 2 online surveys were performed as a first step in gathering user expectations of envisioned garment- mediated emotional communication using ‘warm touch’.

Related Work

Foundational work demonstrating touch is capable of communicating distinct emotions was conducted by Hertenstein et al. [68], [134]. Through 2 direct touch communication studies with encoding-decoding tasks amongst strangers (whereby a ‘sender’ communicates an emotion— through touch alone—to a ‘receiver’; and the ‘receiver’ has to determine the emotion conveyed), discrete emotions (namely, Ekman’s basic emotions involving anger, fear, happiness, sadness, disgust, and surprise, as well as prosocial emotions including love, gratitude, sympathy) were shown to be communicable at 48-83% recognition rates. In the first study, touch emotional communication was limited to the arm while the second study involved anywhere on the body (within reason) to better approximate how touch is used during naturalistic communication settings. Both studies yielded performance measures demonstrating that emotions can be decoded at a rate higher than chance levels and provided insights into tactile behaviors relating to the investigated 99 emotions. For instance, anger was associated with hitting and squeezing, while sadness involved stroking and squeezing [68]. Further, the involvement of more body locations showed that additional emotions (such as happiness and sadness) can be more reliability communicated compared to when a location restriction (i.e., the arm) was implemented [134].

Recent studies focused more on the physical properties of affective touch. Huisman et al. used a wearable sleeve to extract the various tactile expressions and behavior characteristics of 8 different emotions [37], [175]. For example, some emotions (fear, happiness, anger) resulted in significantly larger surface areas usage than others (gratitude, sympathy, sadness). To remove potential confounds from wearable sensing systems since the touch interaction is no longer directly on a user’s skin (such as the sensing sleeve by Huisman et al.), Hauser et al. [28] devised a tracking system involving infrared video and electromagnetic tracking to measure contact characteristics during touch communication of 6 different emotions and found that combinations of the contact characteristics can distinguish various emotions [28]. However, these investigations were only limited to the arm.

Given the success of communicating emotions through direct touch, it comes as no surprise that there is interest in using haptic devices to mediate emotion communication. Using similar encoding-decoding study paradigms, Smith and McLean studied the communication of 4 emotions (angry, delighted, relaxed, unhappy) through a haptic turning knob and found success in 54% of trials [31]. Similarly, Bailenson et al. mediated emotion conveyance through a 2 DOF joystick [32], and found success significantly above chance levels (33%). While these studies show promise in the use of mediated devices to communicate emotions, they are generally limited to force feedback devices (versus wearable systems).

Wearable haptic garments can take many forms, but here we only focus on upper-body garments capable of providing warmth or compression stimuli. Perhaps the most common use of compression is in the context of communicating general positive affect, typically aiming at strengthening personal relationships through emulated touch, in particular, replicating hugs. Some examples include, Mueller et al.’s Hug Over a Distance vest for couples separated by distance [72], Teh et al.’s Huggy Pajama to promote parent-child communication [74], [102], and Cha et al.’s HugMe haptic teleconferencing system to convey intimacy and affection [39]. However, these haptic garments only focus on positive affect, ignoring a wide range of other emotions that may be of interest/use. Some aim to extend to other emotions, albeit for different purposes; Tsetserukou et al. 100 introduced the iFeel_IM! system that uses haptic devices (thermal, pressure, vibration) on various body regions, to enhance and induce emotions (joy, sadness, anger, fear) during online conversations [126], [176]. In contrast, Arafsha et al. described a haptic jacket with heat and vibrotactile actuators to represent 6 emotions (love, joy, surprise, anger, sadness fear) to augment videos for increased emotional immersion [40], [63]. More recently, Delazio et al.’s jacket [17] and Zhu et al.’s pneumatic sleeve [38] delivers pressure and vibrations through a library of feel effects for enhancing AR/VR experiences. However, none of these haptic garments have directly focused on ‘warm touch’ sensations for the use of emotional communication; even when present (e.g., Tsetserukou et al.’s iFeel_IM!), the warmth and compression sensations are decoupled (i.e., they were present on different body regions).

As introduced, SMAs—capable of providing warmth and compressive forces in a single actuation—are an interesting new way of providing mediated haptics. However, since not all haptic modalities are suited to all types of emotional communication [177], [178], there is a need for scoping and bounding the design space specifically for garment-mediated ‘warm touch’ for the communication of emotions. However, effective communication of emotion through mediated touch is currently not well understood [45]. Jones and Yarbrough studied the meaning people assign to touch in everyday interactions [99], yet we cannot assume these translate to mediated touch. In pursuit of better understanding of the haptic strategies used to communicate emotions, investigating the associated mental models and cognitive frameworks that guide these strategies may help in establishing a common reference frame for communication, toward the goal of haptic experiences that are more understandable, memorable and enjoyable.

There are many ways emotional communication can be studied; broadly, the 2 general approaches to classifying emotions are (1) dimensional theory of emotions, mapping emotions as combinations of dimensions consisting of valence (unpleasant to pleasant), arousal (passive/calm to active/excited), and dominance (dominated to dominant) [173]. In the context of researching how haptics relate to emotions, a haptic stimulus can be presented, and users are asked to rate the stimulus in relation to the aforementioned dimensions on bipolar scales. The subjective ratings can then be mapped onto a 2D valence-arousal Circumplex [37], [44] model or a vector model [46] to understand the range of emotions that can be conveyed given varying stimuli qualities. In contrast, another way of studying emotions is (2) the differential theory of emotions, with emotions seen as distinct categories (e.g., happiness, anger, fear) [179], [180]. This study only focuses on the latter

101

(categorical model), since it is the framework that has been employed in the empirical studies involved in direct (human-human) touch communication [68], [134], supported by Nummenmaa and Volynets et al.’s [181] [182] findings of culturally universal representational mapping between body activation patterns and discrete emotions.

In this research effort of scoping and mapping the design space for ‘warm touch’-driven garment- mediated emotional communication, we address the following knowledge gaps: (1) to gain an understanding of the user expectations, strategies, and mental models that users might employ in the act of communicating emotions using a ‘warm touch’ haptic mediated garment, and (2) if those strategies and mental models will change depending on whether users are trying to communicate the idea of an emotion or the feeling of an emotion. The following section describes the two online surveys deployed to investigate each of the aforementioned research gaps to provide user-driven insights, informing intuitive/consistent garment design approaches for emotional communication.

Survey Study 1: Gathering ‘Warm Touch’ Strategies and Mental Models Employed in Emotional Communication

5.2.1. Methods

Online Survey Implementation Methodology

The study setup was inspired by encoding-decoding tasks employed by [28], [29], [68], [134] but performed through a survey. For the scope of this project, we are primarily interested in how and why users choose to communicate various emotions through a mediated affective garment capable of delivering ‘warm touch’. Fifty-three participants (26F/27M, mode=25-34 yrs. old) were recruited through Amazon MTurk and compensated with $8; the survey was limited to those currently residing in the United States, aged 18 and above (more demographic information is presented in Appendix C). Participants were told the project aimed to develop smart clothing that can represent emotions physically and the smart clothing contains technology that replicates human touch, involving squeeze sensations and slight warmth (termed ‘warm touch’ in the survey). They were then shown various emotions (randomized order, 1 emotion per page), and asked to select ‘warm touch’ parameters (body locations, intensity, pattern) corresponding to each emotion. Participants were told the ‘warm touch’ features selected will be delivered all at once through the

102 haptic garment. Figure 5-1 shows a representative survey interface; Figure 5-1 (a) demonstrates the interface where users select ‘warm touch’ on corresponding body locations (torso, shoulders, upper arms, hands, neck), intensity (low vs. high), and pattern (dynamic vs. constant forces with varying speed)—these parameters were based on what is possible given the SMA technology. Figure 5-1 (b) shows a reference chart provided to participants, so that they are grounded with a reference point. Note that participants were allowed to select ‘No Touch’ and ‘None of the Above’ options and provided space to specify alternatives. Along with the presentation of each emotion, prompts were also included to limit the interpretation of the various emotions, as demonstrated in Table 5-1 (adapted from [29]; prompts were designed to ensure all scenarios were emotions that the ‘sender’ is experiencing). The emotions selected involved 4 of Ekman’s basic emotions (Sadness, Happiness, Anger, Fear), 2 prosocial emotions (Gratitude, Love), as well as emotions that people are generally interested in relation to touch communication (Calm, Attention; the latter serving as a reference measure). In addition, participants were told that all communication should be directed to a friend, so that the relationship context was standardized.

More importantly, we probed participants on their thought process behind each ‘warm touch’ haptic strategy decisions for the emotions (they were asked if the selected parameters reminded them of something), as well as other ideas that might better communicate the emotions (i.e., if they had unlimited access to any form of haptic signals, what other strategies would they use instead). These were both presented as required open-ended questions. For each emotion, participants were also asked how confident they were that their designed ‘warm touch’ strategies can communicate each emotion should the ‘warm touch’ be implemented towards their friend wearing a smart garment. This was evaluated based on a bipolar scale from 1 to 5 (not at all to very confident), intended to assess the ease/difficulty in assigning ‘warm touch’ strategies for each emotion.

Table 5-1: Prompts for Emotional Communication Task Emotion Prompt Calm You just finished a calming meditation session and want to share that with your friend. Try to express CALM through ‘warm touch’. Sadness Your pet just passed away. You are feeling sad and want to let your friend know. Try to express SADNESS through ‘warm touch’. Love Think of all the wonderful qualities your friend has. Try to express LOVE through ‘warm touch’. Happiness You just received great news—you landed your dream job. You are feeling really happy and want to let your friend know. Try to express HAPPINESS through ‘warm touch’. Gratitude Your friend just helped you solved a problem. Try to express GRATITUDE through ‘warm touch’. Anger Your friend has done something careless that has set off a series of inconveniences and you are extremely angry. Try to express ANGER through ‘warm touch’. Fear You are extremely fearful of spiders. You just saw a spider on your bed and want to let your friend know you are afraid. Try to express FEAR through ‘warm touch’. Attention You just found something interesting on the internet and want to relay it to your friend. Try to get their ATTENTION through ‘warm touch’.

Data Analysis

A mix of qualitative and quantitative data analysis methods were employed. ‘Warm touch’ haptic strategies per emotion were visualized with parallel sets plots. To determine the relationship between each variable and emotion, χ2 tests for independence were performed and Pearson residuals were calculated to determine the strength of associations. However, since participants were allowed to select multiple locations for each emotion, the frequency of combinations of various body locations were visualized using a bubble plot; the relative color and size of each bubble represents the frequency of participants’ selection for each combination given different emotions. In

104 determining differences between participants’ confidence in ‘warm touch’ for each emotion, Kruskal-Wallis non-parametric tests were used since the data was on ordinal scale and normality assumptions were not met through Shapiro-Wilk test.

All open-ended, qualitative responses were organized using a visual collaborative tool, and iteratively coded to extract meaning within the data. Thematic analysis strategies were employed to explore and reveal themes, as well as using affinity diagramming strategies to spatially cluster coded items based on shared meanings. The spatial clustering and organization process were performed iteratively, merging or refining codes as necessary to identify and extract core themes that provide categorical comparison between emotions. The generated themes supplemented the aforementioned quantitative data visualization and statistical analyses.

5.2.2. Results: Quantitative Survey Data

The relationship between ‘warm touch’ parameter dimensions for each emotion were visualized as parallel set plots. Due to the excessive number of plots, we only present a representative visual (Figure 5-2) indicating how this type of data visualization strategy provides a ‘big-picture’ view of the relationships between parameter dimensions; a full set of plots can be found in Appendix D.

From the χ2 test of independence, each of the parameters (body location, intensity, pattern) were found to be statistically significantly associated with different emotions: (a) Body Location: χ2 (28, n=53)= 46.2, p= 0.017; (b) Intensity: χ2 (7, n=53)= 213.73, p<0.001; (c) Pattern: χ2 (21, n=53)= 105

397.84, p<0.001). Pearson residuals were calculated and presented in Figure 5-3 (a)-(c) showing how positive (blue) and negative (red) the parameters were associated with each emotion, and each variables’ contribution to χ2 statistic.

Comparing between emotions, for body location (Figure 5-3a), the torso was most associated with Happiness and Love and least with Gratitude and Attention. The shoulders were more associated with Calm, Gratitude, Attention, and least with Fear, Anger, Love. Upper arms were moderately associated with Happiness and Attention, while the hands were extremely associated with Gratitude. The neck in particular was interesting because it was moderately associated with negative emotions

106

(Sadness, Fear, Anger) and extremely negatively associated with positive emotions (Happiness and Gratitude); yet, slightly associated with Love and Calm.

For intensity (Figure 5-3b), Happiness and Love were unaffected. There were moderate associations between Sadness, Gratitude and low intensity, as well as Attention and high intensity. Fear and Anger were strongly represented by high intensity, and Calm was extremely associated with low intensity. In ‘warm touch’ patterns (Figure 5-3c), for short constant force, we only see slight associations for Calm, Gratitude, and Attention. A long constant force on the other hand, was extremely associated with Love. The results for quick pulsing forces were polarizing; high arousal emotions like Happiness, Fear, Anger, Attention were strongly represented by quick pulsing forces; conversely, was negatively associated with lower arousal emotions like Sadness, Gratitude, Love, and Calm. Slow pulsing forces were more associated with low arousal emotions such as Sadness and Calm. Overall, except Happiness, each emotion has clear associations with the selected ‘warm touch’ patterns.

However, interpreting above results alone may be misleading because users were allowed to select multiple locations; Figure 5-4 depicts possible body location combinations. For Sadness, there was no clear strategy, but it had the most users picking ‘No Touch’. There could be 2 reasons: (1) users feel Sadness cannot be conveyed haptically, or (2) Sadness is difficult to communicate. Kruskal- Wallis test on participants’ confidence for ‘warm touch’ selections showed statistically significant differences between emotions, χ2(7)= 28.66, p<0.01; post-hoc pairwise comparisons with Wilcoxon rank sum test with Bonferroni corrected p-values showed significantly less user-reported confidence in Sadness than all emotions (p < 0.05) except Anger, Gratitude, and Happiness (table containing p-values adjusted with Bonferroni method is presented in Appendix E). For Happiness, all top choices involved torso, shoulders, and upper arms in the selection. For Calm (n=11), Fear (n=16, 30.2%) and Anger (n=15, 28.3%) ‘all locations’ was the top choice. However, for Anger, except the top choice, fewer location combinations were generally selected (distribution skewed towards top portion of the graph, involving only 1-2 locations). The clearest result was Love, n=21 (39.6%) selected ‘all locations’. Gratitude involved Hands the most (n=10). Finally, for Attention, we note several top combinations but no clear results; the only pattern being that most strategies involved either shoulders, upper arms, or hands.

107

5.2.3. Results: Qualitative Open-Ended Responses

Mental Models Used in Construction of ‘Warm Touch’ Haptic Strategies

As mentioned in section 5.2.1, apart from gathering the ‘warm touch parameters used by participants to communicate each of the emotions, open-ended questions probing the reasons participants selected the specific ‘warm touch’ parameters were also included (participants were asked if the selected parameters reminded them of something, as well as other haptic alternatives that might better communicate the emotions); this qualitative portion of the survey was done to better explain participants’ selected quantitative parameters. The gathered open-ended responses were analyzed using affinity diagramming strategies, by iteratively checking a fit of responses with emergent themes. Through our open-ended response analysis, we identified five major themes in mental frameworks used by participants when constructing ‘warm touch’ communication strategies for each emotion: (1) Representation of Body Sensations, (2) Replication of Typical Social Touch Strategies, (3) Metaphorical Representation of Emotions, (4) Symbolic Representation of Physical Actions, and (5) Mimicking Objects or Tasks.

Since the open-ended question was required by all participants, we were able to also gather the frequency counts each of these mental models were used by participants, which would tell us 108 relative proportions the identified mental models were used for each emotion. In general, each user will employ one of the above strategies (although some involved multiple), however, participants do not use a single strategy for all emotions). Table 5-2 presents the summarized frequency of mental models employed to communicate various emotions, and in the following sections, provide examples of the common responses associated with each of the mental frameworks and highlight some observed trends. Note that in Table 5-2, the total for each row may not add up to the total number of participants because some used a combination of mental strategies.

Table 5-2: Survey 1 participants’ mental model strategies used to communicate emotions Emotion Representation Replication Metaphorical Symbolic Mimicking No of Bodily of Typical Representation Representation of Objects or Touch Sensations Social Touch of Emotions Self-Expressing Tasks Strategies Physical Actions Sadness 25 18 2 7 - 6 Happiness 18 31 3 6 1 - Fear 31 14 3 4 - 1 Anger 25 22 2 4 5 - Love 10 37 9 - - - Gratitude 5 48 1 - - 1 Calm 16 11 9 - 16 1 Attention 8 45 - - - -

Representation of Body Sensations

Participants employing this cognitive framework described bodily changes given the experience of emotions; most commonly in Fear, Anger, and Sadness. For Fear, this was the dominant strategy; out of the 31 subjects, n=10 involved increased pulses on the neck/chest, n=10 spoke of quick pulsing as shivering, n=7 described ‘tensed shoulders/neck’, ‘knots in the stomach’ as physiological manifestations of Fear. In many areas, the reports for Anger were similar to Fear, e.g., n=11 spoke of tensing of neck/shoulders, and many used intense pulsing sensations to represent blood coursing through the body or beating of the heart, hence selected multiple locations (Figure 5-4). However, for Anger, mentions of ‘warmth’ or ‘burning rage’ implied that the concept of warmth is prominent in Anger compared to Fear. Further, descriptions of bodily sensations for Anger and Fear trended towards dynamic sensations; also reflected in the quantitative data (Figure 5-3c, as strongly associated with quick, pulsing forces).

Sadness, on the other hand, had participants describing body sensations on the neck as feeling ‘a lump in the throat’ (n=9) or ‘shoulders/neck slumping or heavy with tension’ (n=6). When torso was involved, it was mostly feeling sick at the pit of their stomach (n=7). For Happiness, Love, and 109

Calm, bodily sensations did not occupy a majority of responses, but whole-body warmth was prominently mentioned (nhappiness=9; nlove=7; ncalm=10). For Love and Calm, the warm feelings were described as being more stationary/constant, e.g., ‘a warm glow that fills you up’. But in contrast, descriptions of warmth for Happiness were more associated with dynamic sensations like ‘butterflies in stomach’ and ‘heartbeats’. Importantly, many body sensation descriptions for these emotions were similar and all three of them are associated with positivity. The emotions that utilize this mental model the least are Gratitude and Attention; for Gratitude, the participants who employed this strategy of representing body sensations based it on warm, enveloping feelings in the torso when feeling appreciative, while Attention was described as the ‘body posture in alarm’ or ‘feeling attentive’.

Replication of Typical Social Touch Strategies

Gratitude and Attention involved this mental model the most. For Gratitude, descriptions involved ‘handshake’ (n=15), ‘hug/embrace’ (n=16), ‘pat on the back/shoulders’ (n=10), all of which are common appreciation expressions. Mental models for Attention were mostly ‘tapping the shoulders’ (n=13), ‘grabbing arms’ (n=5), ‘squeezing of the shoulders/hands’ (n=3), aimed at directing receivers’ focus, hence the utilization of quick pulsing forces (Figure 5-3c). Surprisingly, a non- negligible portion of users (n=10) selected all locations to present ‘warm touch’ for Attention; for this group, 70% of them equated more surface area to larger attention capture. This mental strategy was also popular for Love and Happiness. For Love, unsurprisingly, ‘hugs’ were most cited (n=28) and we observe the use of many body locations; n=21 opted for ‘warm touch’ on all locations (Figure 5-4). Similarly, in Happiness, many chose to communicate Happiness with ‘a hug’ (n=17), or positive expressions like ‘shoulder pats’. Again, descriptions for Love tended to be less dynamic than Happiness. Anger mostly described actions typically done in anger (e.g., ‘forceful grabbing/ squeezing’) on the hands, arms, or shoulders. In contrast, users’ mental model for Sadness mostly concerned supportive touches, commonly ‘a hug’ (torso/shoulders) for consolation (n=10), ‘holding/squeezing the hands for comfort’ (n=8), or ‘arm around shoulders’ (n=5). The emotions that utilize this mental modal the least were Calm and Fear. The descriptions for Calm were similar to Sadness in that most ‘warm touch’ were aimed at providing reassurance. As for Fear, all accounts were related to activating the startle response (i.e., warning/alarm) (n=9) or reflexive actions e.g., ‘grabbing someone in shock’ (n=5).

110

Metaphorical Representation of Emotions

A small subset of participants chose to use metaphors to describe the emotions and associated them with various ‘warm touch’; in this case, they did not provide examples but rather, connected attributes of emotions such as valence/arousal, spatial descriptions (e.g., up/down) to perceptual properties of ‘warm touch’ sensations such as speed/intensity, or even descriptive metaphors such as comparing emotions to nature. For instance, one participant commented that Calm was like ‘a sunlight feeling of warmth, something that spreads across your entire body that makes you feel at peace’ and chose long, constant force on all locations. The use of metaphorical representations was more common in Love and Calm, typically relating to the enveloping ‘fuzzy’ feelings or power the emotion brings (e.g., ‘Love is an intense, consuming head to toe emotion that fills you’). Some other examples include ‘feelings of heaviness’, ‘downward movements’ to describe the thought- process for Sadness and ‘high intensity jolts’ for Anger. However, overall, this mental strategy was not commonly used and when employed, such as for emotions like Love and Calm, representations tended to overlap with other cognitive frameworks, especially those relating to body sensations.

Symbolic Representation of Self-Expressing Physical Actions

Some participants also symbolically represented physical actions that are self-expressing given an emotional experience (we differentiate these from social touch whereby these self-expression actions are not in contact with another person, but rather are self-performed). For instance, for Sadness, one might wring their hands or cry, participants in turn activated ‘warm touch’ sensations on associated body locations to symbolically represent the expressed actions (e.g., constant forces on the hands to represent wringing of hands; pulsing on the neck to represent crying/sobbing). For Happiness (n=6), multiple body locations were activated to symbolize dancing or jumping with joy; Fear (n=4) saw several participants activating ‘warm touch’ on the hands to symbolize seeking help from someone else; and Anger (n=4) all activated the hands to symbolize clenching of the fists.

Mimicking Objects or Tasks

Finally, mimicking physical actions or tasks was another theme observed, albeit to a much lesser degree. It is most commonly used to describe the ‘warm touch’ strategies for Calm (n=16). Out of the 16 responses, 13 of them involve task-based situations (in which the ‘warm touch’ sensations provided accomplishes a task in itself, distinguished from regular social touch because there is a clear task-related goal). In this context, 13 of the responses involved using the ‘warm touch’ as a 111 form of massage for the release of muscle tension; this always employs pulsing forces on either the torso or shoulders. The other 3 responses involved mimicking other objects that can provide calming sensations including compression jackets or weighted blankets. A small subset of participants also used this strategy for Anger, in particular, those who mimicked ‘strangling’/ ‘choking’ actions (n=5)—typically presented on the neck; we also distinguish this from typical social strategies because such aggressive actions are atypical in everyday interactions [99].

No Touch

We know from past research and even our personal, everyday experiences that there are emotions that are less associated with touch than others. Here, Sadness resulted in the most ‘No Touch’ option (n=6). Some participants thought Sadness was about feeling hollow and empty so any touch would be contradictory (n=3), while others felt that Sadness could not be easily communicated via ‘warm touch’ haptic sensations. Other emotions where participants selected ‘No Touch’ options included Fear, Gratitude, and Calm (n=1).

Alternative Strategies Proposed by Users and Other General Comments

To mitigate tendencies for participants to force matches between emotions and ‘warm touch’, we encouraged considerations of alternate forms of haptics (i.e. no longer limited to ‘warm touch’). The summarized top strategies are presented in Table 5-3.

Table 5-3: Alternative haptic strategies proposed by Survey 1 participants Emotion Alternative Strategies/Ideas Sadness Cold sensations (n=22), Involve the face (n=4), Irregular pulses (n=3), Color-Blue (n=1) Happiness Vibrations (n=6), Rhythmical pulsing (n=6), Color- Yellow (n=1) Fear Cold sensations (n=28), Vibration- Shiver (n=8), Rapid Heartbeats (n=3) Anger Sharp sensations/Vibration (n=13), Cold (n=8), Increase heat (n=4), Abrasive materials (n=1) Love Caress (n=10), Rippling/Flow (n=4), Therapeutic Massage (n=4), Color-Pink (n=2) Gratitude Patting (n=8), Warm Glow all over the body (n=8), Vibrations (n=7) Calm Caress (n=8), Coolness (n=5), Vibrations (n=4), Audio cues (n=3), Color- Blue (n=1) Attention Vibration (n=12), Audio cue (n=4)

From Table 5-3, the most requested addition (by 42-53% users) was the use of cold for Sadness and Fear. For Sadness, cold was said to better convey emptiness, sense of unease, or illusion of moisture from crying; for Fear, cold was selected as chills, a sense of dread, or blood draining from

112 the body. Interestingly, we saw several users (n=8) who thought cold would be better suited for Anger instead of warmth; in contrast to the majority who thought that warmth was ideal for this emotion. Another common idea was vibrations for: Happiness (e.g., high energy everywhere), Fear (e.g., with cold for chills), Anger (e.g., shaking/jabbing), Gratitude (e.g., near the heart), Calm (e.g., massage), and unsurprisingly, Attention (e.g., similar to mobile notifications). We also noted some who wanted ‘warm touch’ coupled with other modalities, e.g., audio or color changing interfaces.

5.2.4. Survey 1: Summary

From Survey 1, several observations can be made. First, the mental models used were related to the body locations on which ‘warm touch’ was employed on. Intimate areas (torso/neck) were more likely used to describe bodily sensations than the extremities; except torso in replicating hugs and the neck for strangling actions. The shoulders had a mixed role, e.g., to demonstrate muscle tension, or as an area for social interactions like hugs/showing support. The hands and arms being non- vulnerable body areas, were commonly used as handshakes or in grabbing attention. The ‘warm touch’ intensity and patterns better aligned with qualities of emotions such as valence/arousal.

Secondly, representation of body sensations was more likely to involve negative, basic emotions (Sadness, Fear) than prosocial ones (Love, Gratitude) or Attention, likely due to changes in physiological responses (e.g., breath/heart rate) associated with basic emotional encounters. The mental model for Anger was split between bodily sensations and social touch; Anger has strong body activations and is typically expressive in nature. Broadly, with the exception of Gratitude and Attention, the mental models were not ‘clear cut’, and in many cases, were derived from differing intentions with the conveyance of a given haptic profile (e.g., was the sender attempting to make the receiver feel a specific body-touch that normally cause an emotional experience, or were they trying to replicate the visceral experience of an emotion directly?); this could be problematic if the receiver of the ‘warm touch’ employ a mental model that is in conflict with what the sender has in mind. Currently, the top 2 strategies used to communicate emotions are replication of body sensations and social touch strategies, but Survey 1 results did not account for variabilities in strategy selection. There is a possibility that this survey resulted in different interpretations. Participants were told to communicate emotions, but in Table 5-1’s prompts, the word ‘express’ may have resulted in ‘expressing/representing’ emotions instead of ‘communicating/conveying’ them. Further, one of the main goals of this study was to determine ‘warm touch’ strategies used to communicate emotions. While we observed general trends, since 113 the number of possible variable selections results in many combinations, many users opted for 4-5 or all locations to communicate most emotions (Figure 5-4, Note the concentration of counts at Figure 5-4’s last row; many opted to present ‘warm touch’ on all body locations. In some cases even, multiple mental models were used for a single emotion, e.g., combining body sensations and social touch strategies), which made it difficult to determine saliency.

Hence, we performed a follow-up survey, specifically targeting these limitations. To address possible varied survey interpretation, we distinguished between communicating versus eliciting emotions. Two objectives were selected (1) to clarify the difference in our first study, and (2) to clarify the difference for future research. They are rarely delineated in current literature when designing affective garments and could have far-reaching impacts in terms of application space. Secondly, we limited the selection of body location to a single option to tease out the most salient features of ‘warm touch’.

Survey 2: Investigating If Communication Objective Changes User’s Mental Models In ‘Warm Touch’ Haptic Communication

5.3.1. Methods

Online Survey Implementation

A total of 80 participants (residing in United States, aged >18) were recruited through Amazon MTurk, and randomly divided into 2 groups; Group 1-Communicate (n=40, 21F/18M/1 preferred not to answer, mode= 25-34 yrs. old) and Group 2-Elicit (n=40, 18F/22M, mode= 25-34, 35-44 yrs. Old, more information can be found in Appendix C). The survey was identical in both groups, with the exception that

Group 1 (Communication) was told that their task was to communicate how they felt to their friend based on the given emotions. (‘A friend of yours is wearing one of these smart garments and you are physically separated from each other. Your task is to use the smart garment to let your friend know your emotional state (i.e., communicate how you feel))’; versus

Group 2 (Elicitation) who were told their task was to make their friend feel certain emotions (‘A friend of yours is wearing one of these smart garments and you are physically separated from each

114 other. Your task is to use the smart garment to elicit certain emotions in your friend (i.e., make your friend feel certain emotions)’).

The order of emotions was randomized, and subjects were asked to select parameters (location, intensity, pattern) for each emotion; this time, multiple body locations were not allowed. Again, we probed participants on their thought process behind the ‘warm touch’ decisions as well as the option to voice alternatives. We also eliminated the use of prompts due to difficulties in synchronizing scenarios between communicate/elicit groups. Further, ‘Attention’ was removed since it wouldn’t fit into the communicate/elicit narratives and has the added benefit of reduced survey length.

Data Analysis

Similar to Survey 1, the data was first visualized with parallel set plots to show distributions and associations between the parameters. χ2 test of independence was performed to determine association between parameter dimensions (body location, intensity, pattern) and the different emotions. Since there are now two groups, comparison between groups for each location, pattern, and intensity variable was also performed with Fisher’s Exact test (since some cell counts were <5.) Since participants are now asked to only select one option for each parameter, the parameter combinations' frequency counts can be tallied (presented as bubble plots). As before, along with each ‘warm touch’ design strategy, participants were asked to provide mental models used during the process—all such qualitative responses were iteratively coded using affinity diagramming strategies as before, this time guided by the categories extracted from Survey 1.

5.3.2. Results: Quantitative Survey Data

The relationships between ‘warm touch’ parameter dimension categories were visualized with parallel sets plots and supplemented with statistical data. Within each of the ‘communicate’ and ‘elicit’ groups, each of the three parameters (body location, intensity, pattern) were found to be statistically significantly associated (by χ2-squared test of independence) with the different emotions, p <0.001. Figure 5-5 demonstrates the strength of association for each variable. Comparing between ‘communicate’ and ‘elicit’ groups, the Fisher’s exact test results are presented in Table 5-4. We saw statistically significant differences (p <0.05) between groups for Fear in body location, and Happiness in intensity. Only plots for Happiness (Figure 5-6) and Fear (Figure 5-7) are shown; all others are in Appendix F.

115

Group 1: Communicate Group 2: Elicit

* *

Body Location

* * Intensity

Pattern

Table 5-4: Survey 2’s Fisher's Exact Test for comparison between groups

Fisher's Exact Test Results p-value Between Groups Location Intensity Pattern Sadness 0.7363 1 0.174 Happiness 0.1331 0.0435* 0.1043 Fear 0.00193 * 0.7555 0.1171 Anger 0.1089 1 0.5429 Love 0.4632 0.6541 0.9168 Gratitude 0.7365 0.8116 0.9376 Calm 0.9059 0.08718 0.332

Overall, the ‘warm touch’ parameter trends were similar between Survey 1 and Survey 2), demonstrating the successful replication of Survey 1’s results (i.e., if both groups in Survey 2 were combined, the distributions are similar to Survey 1). However, when Survey 2 was split into ‘communicate’ and ‘elicit’ groups, we noted some differences. First, we briefly present results for 116 emotions where there were no statistically significant differences between groups but the ‘warm touch’ strategy trends (i.e., salient variables) for each emotion are still relevant. For Sadness, low intensity (>70% users selected this) and long, constant force were top choices. Anger was characterized by high intensity (>95%), quick pulsing forces, with the upper arms and neck as top locations. Results for Love across groups had torso and a long constant force as the most selected

(upper arms were the least used; nGroup1=0; nGroup2=2). Like Survey 1, it had no difference between intensities. For Gratitude, the shoulders and hands were the majority at low intensity, short constant force. This is slightly different from Survey 1 which had prominent hand usage (77% of Survey 1’s users used hands to communicate Gratitude; shoulders were the 2nd top choice, so major trends are consistent). The final emotion is Calm; we saw a majority (83-95%) of the selection involved low intensity sensations, with shoulders as the top choice. Similar to Survey 1, both groups had either slow pulsing forces or a long constant force as the top pattern.

Figure 5-6: Parallel Set Plots for Happiness. Left: Group 1 (Communicate); Right: Group 2 (Elicit)

Figure 5-7: Parallel Set Plots for Fear. Left: Group 1 (Communicate); Right: Group 2 (Elicit)

There were significant differences between groups for some parameters. For Happiness in the intensity parameter (p=0.04), majority (64%) of participants selected high intensity in Group 1 (Figure 5-6, left) but more participants (63%) selected low intensity in Group 2 (Figure 5-6, right). 117

One big contributor to this difference is the association between patterns and intensity. More Group 1 users selected quick pulsing forces; within that, 16/25 of them paired this pattern to high intensity. In contrast, Group 2 had more users selecting long constant force, more commonly paired with low intensity. On the other hand, for Fear (Figure 5-7) both groups used high intensity (82-87%). But for Group 1, quick pulsing forces were chosen=28, 72%) while for Group 2, choices were split between quick pulsing forces (n=17, 44%) and long constant force (n=15, 38%). Comparing body locations, Group 1 had an even distribution in the use of upper arms, hands, and neck. Group 2 had a strong skew towards the neck to elicit emotions that was statistically different (p=0.001) between groups. After understanding the trends for each emotion given the different parameters, we can also extract top strategies (i.e., combination of parameters most frequently used by participants), presented in Figure 5-8.

Figure 5-8: Balloon Plot of Selection Frequency for All Possible ‘Warm Touch’ Combinations. Left: Group 1- Communicate; Right: Group 2- Elicit

118

Here we highlight strategies that have at least 7 participants (>17.5%, indicated as ‘orange-yellow’ in Figure 5-8) agreeing on the specific combination. For Fear, interestingly, both groups’ top strategies are drastically different; ‘Hands + Quick, Pulsing Forces + High Intensity’ was a top choice in Group 1 (n=9) but Group 2’s choice was ‘Neck + A Long, Constant Force + High Intensity’ (n=10). The second top choice was the same across groups, ‘Neck + Quick, Pulsing Forces + High Intensity’. There were some emotions that had top strategy agreements between groups. For instance, both groups’ top strategy for Anger was ‘Upper Arms + Quick, Pulsing Forces

+ High Intensity’ (Group 1, Group 2: n1=9, n2=5). For Love it was ‘Torso + A Long, Constant

Force + High Intensity’ (n1=9, n2=7) as the top strategy, followed by ‘Torso + A Long, Constant

Force + Low Intensity’ (n1=8, n2=7). Calm’s top two strategies were also consistent: ‘Shoulders +

Slow, Pulsing Forces + Low Intensity’ (n1=8, n2=7), followed by ‘Shoulders + A Long, Constant

Force + Low Intensity’ (n1=5, n2=4).

5.3.3. Results: Qualitative Open-Ended Responses

Mental Models Used in Construction of ‘Warm Touch’ Strategies for Each Emotion

Themes in participants’ mental frameworks relating to ‘warm touch’ strategies for each emotion were identified; definitions for each of the 5 major categories can be found in Survey 1. Here we take a step further, breaking down social touch strategies into subsequent categories: (a) Support— for reassurance (e.g., ‘being there for someone’), (b) Reflex—involve reflexive actions (e.g., grabbing in response to something scary), (c) Attention—directing focus to the initiator, (d) Affection—positive regard, (e) Control—exertion of dominance (e.g., ‘grabbing and shaking them’, ‘let someone know how strong I am’), (f) Appreciation—expression of gratitude (e.g., ‘saying thank you’), (g) Neutral Presence (e.g., ‘just being there’—with no explicit verbalization in regards to a purpose). Comments that could not be categorized (no derived meaning relating to mental models) were categorized as ‘Other’. Table 5-5 (Group 1: Communicate) and Table 5-6 (Group 2: Elicit) present the frequency tallies for each mental model strategy and we discuss the results per emotion, comparing between groups.

119

Table 5-5: Survey 2 Group 1’s (Communicate) mental models used during‘warm touch’ strategy selection GROUP 1: COMMUNICATE

Replication of Typical Social Touch Strategies

Self- Mimicry

Metaphorical Bodily Expressing of Emotion Representation Other Sensations Physical Objects

of Emotions

Reflex Neutral Neutral

Control Actions or Tasks

Support

Presence

Affection

Attention Appreciation Sadness 18 8 1 4 - - - - 11 - - 4 Happiness 9 1 2 - 11 - - - 4 2 9 4 Fear 20 1 8 1 - - - - 3 3 4 4 Anger 15 - - 3 - 10 - - 2 3 4 4 Love 8 1 - - 24 - 1 - 2 - 1 5 Gratitude 6 2 - - - - 25 - 2 - 2 4 Calm 11 4 - - - - - 6 5 - 15 3 .

Table 5-6: Survey 2 Group 2’s (Elicit) mental models used during‘warm touch’ strategy selection GROUP 2: ELICIT

Replication of Typical Social Touch Strategies

Self- Mimicry

Metaphorical Bodily Expressing of Emotion Representation Other Sensations Physical Objects

of Emotions Reflex

Neutral Neutral Actions or Tasks

Control

Support

Presence

Affection

Attention Appreciation

Sadness 12 17 - - 2 - - - 4 1 3 3 Happiness 8 - 5 - 21 - - - 2 - 5 1 Fear 16 2 3 - - 1 - - 3 - 14 2 Anger 13 - - - - 8 - - 6 4 7 2 Love 4 - - - 32 - - - 1 - 1 4 Gratitude 3 2 - - - - 30 - 2 1 4 1 Calm 5 6 - - 2 - - 9 2 6 10 3

First, we checked if our results from Survey 1 were replicated (i.e., if we combined Groups 1 and 2’s results for Survey 2, the overall trends should be similar to Survey 1 with some exceptions, which would help explain how the use of the language ‘communicate’ and ‘elicit’ might change one’s mental model). Overall, we found major consistencies with Survey 1; body sensations and social touch remained top mental models. Similar to Survey 1, Gratitude, Love, and Happiness had majority of subjects opting for social touch strategy as their mental model. On the other hand, the mental model usage distribution between body sensations and social touch strategy were relatively uniform for Sadness, Anger, and Calm (when the categories are collapsed). Fear on the other hand, like in Survey 1, skewed towards bodily sensations compared to other mental models. Next, we

120 discuss differences and added knowledge from the Group 1 and 2 comparisons.

For Sadness, twice as many users from Group 2 (Table 5-6) used the mental model of supportive social touch compared to Group 1 (Table 5-5) who mainly used body sensations or metaphorical representations; this was surprising—we expected Group 2 (instructed to elicit emotions) to opt to mimic body sensations so recipients could relate to the felt ‘warm touch’. One explanation is that participants may only be concerned if the emotions originated within themselves or others. Hence, when Group 1 was told to communicate how they were feeling, they might only associate it with emotions originating from themselves. When Group 2 was told to make someone else sad (i.e., elicit), the act is uncommon, hence users may default to typical strategies like providing support— even though this will likely be unsuccessful in eliciting emotion. We also notice more participants (Group 1) used metaphorical representation to communicate Sadness than other emotions, likely because Sadness is an emotion that has stumped our participants.

Next, for Happiness, Group 1’s major strategies were split between affectionate social touch, body sensations, and mimicry. In contrast, Group 2 had a clear majority (>50%) using affectionate social touch. One reason is that Group 1 may have more options in ‘communicating’ Happiness compared to Group 2. Outside of hugging, there may not be many alternatives to elicit Happiness. In line with this, more users in Group 1 used mimicry (e.g., dancing, high five—that are more expressive) than Group 2. These reflect the quantitative data; majority of Group 1 chose high intensity (to show expressiveness), but the trend was reversed in Group 2, with more opting for low intensity (for affectionate touch). For Fear, the major difference was greater use of mimicry by Group 2 participants than Group 1, specifically, mimicking the act of choking/strangulation (distinguished from reflexive touch because strangulation is atypical in social touch). In fact, 13/14 participants who selected mimicry in Group 2 spoke of strangulation on the Neck to elicit feelings of Fear (in Group 1 only n=1 spoke of strangling acts). This could be because an unexpected sensation on the neck can presumably incite Fear (Group 2), but when told to communicate feeling Fear (Group 1), the origin of Fear isn’t necessarily caused by tightening of the neck. Instead in Group 1, we see more use of body sensations and reflexive touch, common to internal feelings or automatic responses to feeling Fear.

As for Calm, there were more (2-fold) utilization of body sensations in Group 1 than Group 2 but it was not the top category; the most popular being the mimicry of tasks. The majority using this strategy focused on the shoulders with low intensity, slow pulsing sensations to mimic a shoulder 121 massage. Specific to Group 2, we observed the use of symbolic self-expressing physical actions, where n=6 spoke of eliciting Calm as either a reminder to be ‘centered and calm’ (with constant forces) or similar to meditation techniques, using rhythmic sensations as a focal point (all slow pulsing forces). We likely do not see this in Group 1 because users were told to communicate their own feelings of Calm, so the focus was not about a way to make someone else feel Calm. Neutral presence was a category unique to Calm; described as an unmoving force of ‘just being there’, most typically paired with either short/long constant force. Finally, in Survey 1, Sadness resulted in the most participants selecting ‘No Touch’. In contrast, here in Survey 2, it was Anger (n=3, all originating from Group 1). All participants who opted not to provide touch sensations for Anger thought that it was inappropriate to present ‘warm touch’ to anyone while feeling angry. Other emotions that included ‘No Touch’ included Sadness (n=1, Group 1) and Calm (n=1, Group 2).

Alternative Strategies Proposed by Users and Other General Comments

Results for the optional question (alternate haptic strategies) are shown in Table 5-7. Similar to Survey 1, the most requested addition was cold for Sadness and Fear. We again found a small group of users who thought cold would be better suited for Anger. Another common idea was using vibration for Happiness and Anger. Finally, a major observation was that some users (from both groups) requested multiple locations to be included for several emotions including Happiness, Love, and Gratitude.

Table 5-7: Alternative Haptic Strategies Proposed by Users Divided by Group Emotion Group 1: Communicate Group 2: Elicit Sadness Cold (n=11) Cold (n=15) Happiness Multiple locations (n=4), Vibration (n=5) Multiple locations (n=6), Vibration (n=10) Fear Cold (n=9) Cold (n=14) Anger Vibration (n=8), Intense heat (n=4), Cold (n=1) Vibration (n=6), Intense heat (n=2), Cold (n=4) Love Multiple locations (n=4), Caress (n=5) Multiple locations (n=8), Caress (n=6) Gratitude Multiple locations (n=3), Around heart (n=3) Multiple locations (n=2), Patting back (n=5) Calm Caress (n=4), Audio (n=1) Caress (n=1), Audio (n=1), Waves (n=3), Cold (n=4)

5.3.4. Survey 2: Summary

In Survey 2, we gathered more insight on ‘warm touch’ parameters for each emotion. For intensity and pattern, the resulting overall trends were fairly consistent across both Surveys. However, there were some observed differences in the strength of association between Survey 1 and Survey 2’s 122

Groups 1 and 2 (e.g., for Love, intensity is affected differently in ‘communicate’ and ‘elicit’ groups compared to Survey 1). We hypothesize two reasons: (1) there are inherent differences in attitudes of surveyed participants—since our participant pool is relatively small, some of the differences may be due to sampling variability, and (2) the lack of clear framing in Survey 1 in contrast to Survey 2 (where explicit instructions were provided to distinguish between ‘communicate’ and ‘elicit’) may have contributed to these differences. In other words, the response distribution could have changed based on the prescribed intent. If anything, it informs us that there may be value in adding specificity in instructions. Since these are post hoc survey data, additional study is needed to disambiguate the underlying cause. An adaptive study could be done with actual haptic garments; comparing between ‘communicate’ and elicit’ prompts. With possible observed differences, we can then selectively engage with those participants to probe reasons for their different choices.

The parameter that changed the most was body location—since we only allowed single body location selection in Survey 2, we observed some salient locations for emotions like Fear, Anger, Love, and Calm. However, the drawback is that some emotions may require multiple body locations (Happiness, Love, Gratitude, per open-ended responses)—a general trend seemed to be that for positive emotions, more body surface area was welcomed. Consistent with Survey 1, Sadness was still the most unclear, with no single strategy that stood out. Comparing between ‘communicate’ and ‘elicit’ groups there seem to be clear ‘warm touch’ strategy differences in Fear and Happiness, but not other emotions.

Discussion and Implications

Revisiting the study’s motivation to understand how user strategies and mental models are employed during emotional communication (through a hypothetical affective garment capable of delivering ‘warm touch’ haptic stimuli), two online survey studies were created and deployed. Unique to this study, we not only gathered data on ‘warm touch’ parameters employed by users to communicate emotions, but also obtained qualitative responses relating to reasonings behind the ‘warm touch’ decisions to understand mental models used by participants. Mental models not only impact haptic device design decisions (since we want systems that align with user expectations) but may also influence how haptic signals could be interpreted. Given the results, the implications for affective haptic system designs are vital to distill.

123

One observation from Survey 1 is that when users were allowed multiple location selections, many opted for 4-5 locations to communicate most emotions; some even combined several mental models given multiple locations (i.e., employing redundant haptic/mental strategies for a single emotion). We draw parallels with the concept of ‘choice overload’ [57], [58], whereby people could struggle with decision-making when faced with too many options. With the option to multi-select body locations, intensity and pattern, it resulted in many possible options. Users could have reported all ‘possible’ methods of communication (‘just to be safe’/’just because they can’) rather than a single ‘best’ strategy. This has implications in the design/implementation of affective haptic systems. While a system should allow a certain degree of flexibility for users, with too many possible variable combinations, the emotion being communicated over the haptic garment can get lost. When we limited the selections in Survey 2, however, we found that some emotions will still likely require multiple body locations. There needs to be a balance between the autonomy provided to users and the design space in which users may operate.

When we divided Survey 2 into ‘communicate’ and ‘elicit’ groups, we saw differences in relative perspectives participants assumed for different emotions. Some offered less room for interpretation, for instance when a single action could be used to both communicate and elicit (e.g., a hug can communicate Love and elicit pleasant emotions if sensations are pleasant; it also has strong symbolism, ubiquity and social acceptability) this may be why there is more coherence for positive/prosocial emotions than negative ones. For negative emotions, we noted a greater range of available actions. For example, with Sadness: a user could mirror (1) felt body sensations (make others empathize with one’s Sadness); (2) sensations to be felt (elicit Sadness through ‘heavy’ sensations); (3) support-strategies for negating Sadness one is tasked to elicit/communicate (provide support by giving a hug); or (4) attention seeking actions to indicate one needs help (e.g., squeeze of the hand). ‘Mitigation’ scenarios are understandably rarely present in positive emotions. The perspective that the participant is taking (i.e., the role that one is feeling (egocentric) vs. empathizing (allocentric)) will impact how users deliver and understand ‘warm touch’. If this ‘world-view’ is not in sync between the sender, receiver, and the capabilities of the haptic garment (given a use context), there could be a disconnect.

As a first attempt in understanding how users expect to communicate emotions given a hypothetical garment capable of ‘warm touch’, we observed the complexities of affective dimensions of touch. Meanings of touch may be ambiguous, disparate [51] and contextual [10][38]. Further, the extent

124 to which emotions can be communicated also depends on the haptics afforded by the affective garment. Therefore, the goal perhaps should not be to find a ‘universal warm touch recipe’ for communicating emotions, but instead contemplate how we can creatively design affective garments, such that the system is flexible enough to reconcile differences in users’ mental models but not allow so many choices that the message gets lost. Conversely, an alternative goal could be to communicate to both sender and receiver the system’s operational model (e.g., make system discoverable to aid ‘mental model-building’).

Several options can be considered. First, we could divide ‘warm touch’ by emotion characteristics and intents. For instance, some body regions are more associated with certain emotions than others (e.g., the neck in negative emotions, the torso in positive affect, upper arms for reactive emotions like Anger) and similarly, varying stimulus qualities (pattern, intensity) can be pleasing or bothersome. When coupled with visuo-audio modalities, it might be enough to elicit/communicate different emotional experiences. Another option is to find commonalities between mental strategies (e.g., for Love, ‘warm touch’ on the torso was associated with warm glow but could also convey a hug). Likewise, ‘warm touch’ parameters were similar for some emotions (e.g., Anger and Fear). With an accompanying context in the design, the emotion might be easily identified. All these are possible strategies, but each has inherent pros and cons. For instance, similarities between Anger and Fear is a double-edged sword; if the goal is to distinguish between emotions, it poses a disadvantage. Therefore, (1) the application can guide choices through suggested parameters, (2) or offer mitigating strategies (e.g., calibrating expectations of sender/receiver); or (3) employ tactics used by earlier work and design discrete physical manifestations of the garment to align only with a small subset of emotions (e.g., hug vests), so all parties are restricted in their understanding of what is/can be communicated through the system.

If this garment is to be used for both communication and elicitation applications, however, another alternative is extracting common actions/tasks/body sensations across emotions such that there is relative flexibility in mental model choice, yet the possible design space is narrowed. Table 5-8 demonstrates how this could be done.

125

Table 5-8: Possible design strategy for Affective Garment to be used for Communication and Elicitation Contexts Action/Task Emotions Covered Body Location Intensity Pattern Sadness (Support), Calm (Body sensations) Torso, Shoulders Low Long Constant Hug Love (Affection), Happiness (Affection) Torso, Shoulders High Long Constant Gratitude (Appreciation), Sadness (Support), Holding hands Hands Low Long Constant Love (Affection) Calm (Mimicry, Self-expression), Gratitude Massage Shoulders Low Slow Pulsing (Appreciation) Grabbing arms Fear, Anger (Reflex, Attention) Upper Arms High Short Constant Strangling Fear (Mimicry) Neck High Long Constant Sensations Emotions Covered Body Location Intensity Pattern Sadness (Body sensations- tension) Shoulders, Neck Low Slow Pulsing Happiness (Body sensations- full body Torso, Shoulders, Bodily High Quick Pulsing activation) Upper Arms sensations* Fear, Anger (Body sensations- full body Torso, Shoulders, High Quick Pulsing activation) Upper Arms, Neck *To further differentiate between emotions: (i) Vibrations for Happiness (possibly also Anger), (ii) Cold for Fear (possibly also Sadness)

In particular for bodily sensations, Sadness was selected to be represented by the feeling of tension on the shoulders as well as the constriction felt in the throat, while the representations of Fear, Anger, and Happiness were driven by stronger body activation given these high arousal emotions (also supported by [49][50]). We also incorporated areas where participants suggested alternatives of added cold (Fear, Sadness) and vibrations (Happiness, Anger) could be added to clarify between emotions. Of course, these are shaped based on this survey’s results alone (with added researcher interpretation when consolidating the findings), hence these should be validated in the future and further fine-tuned for specific applications.

The aforementioned possible design strategies were discussed broadly for unspecified applications, assuming all emotions have to be included. As discussed, not all haptic modalities are well-suited for all types of emotional communication; some emotions had more ‘No Touch’ selection than others (e.g., haptic sensations were inappropriate), as well as ‘None of the above’ (e.g., ‘warm touch’ was non-optimal). The two emotions with most selected ‘No Touch’ were Sadness and Anger. For Sadness, some thought touch is contradictory since the emotion is about feeling empty, some do not want to engage others while sad, while others thought Sadness cannot be communicated via touch (Note: for the first reason, there could be a case where one experiences continual haptics, with an absence of a stimulus, be interpreted as an affective communication. Similar to a long pause in a text conversation). For Anger, all who decided on ‘No Touch’ thought

126 it was inappropriate to present haptics at all while one is angry; these comments only originated from Survey 2, Group 1 participants. For some emotions/situations, haptic sensations may not be wanted nor appropriate (application wise, the communication of emotions haptically may occur less in these emotions than others; and there is huge importance of ethics/consent and mapping the boundaries of each). Further, most of the ‘None of the above’ selection reasons have been introduced in the alternative haptic strategy sections (Table 5-3, Table 5-7). We note that consistently across both surveys, the most popular suggested alternatives were cold sensations for Fear/Sadness, and vibrations for Happiness/Anger (to a lesser extent). In particular for Fear and Sadness, since the cold sensation is in direct contrast to the warmth that ‘warm touch’ provides, there is a possibility that ‘warm touch’ just may not be suited for communicating these emotions— further research has to be conducted to validate these findings.

5.4.1. Limitations and Future Work

One area this study is unable to definitively answer is the question of body location. From survey 1, with multiple locations selected, it is unclear if participants meant a given emotion could be conveyed in any of the locations or if the emotion is best conveyed by stimulating all selected locations at once. We attempted to distinguish these in Survey 2, assuming that we might clarify a single ‘best’ location for each emotion. However, because Survey 2 only allowed a single body location, if the best conveyance is through simultaneous locations, then we still do not know that the single-choice location in Survey 2 is the ‘best’. In terms of mental models, this is a first attempt to understanding the ‘whys’ associated with ‘warm touch’ to communicate emotions. Other factors not investigated in this study could affect these findings. For instance, if intensity is now a parameter that could be used to modulate the emotional intensity (e.g., low intensity for a little Anger vs. high intensity for more Anger), the mental models might change. Further research is needed to understand these confounds.

Additionally, while an online survey allowed gathering of general populations’ expectations, the situation is hypothetical; with the use of an actual garment with experienced haptics, the ‘warm touch’ strategies and mental models might vary—a proposed study is presented in Appendix G (the initial intent was to execute the study but due to restrictions imposed by COVID-19, the study has been postponed indefinitely). In addition, evaluation with the garment is needed to distinguish if conveyed emotional information (i.e., the intent) versus actual felt emotion (i.e., if it induces emotion) is possible. Along those lines, we must determine the degree of difference between direct 127 touch versus those mediated by an affective garment; it is conceivable that the idea of haptic clothing may trigger alternate associations and change how ‘warm touch’ communication strategies were devised. Further, in the current survey, participants were likely not thinking of distinguishing between emotions; it is plausible that the results might change if participants were told explicitly to devise unique strategies for each emotion. Moreover, touch is known to be affected by cultural expectations, gender, preferences, personality, and relationship between communication partners [11]. Given the current setup (only participants from the United States were recruited and the exact degree of relationship of the friend/communication partner was not clarified), the external validity may be limited.

Another potential limitation in these studies is the problem of varying participant perspectives in survey interpretation. While several pilots were conducted to minimize confusion in the survey development process, we cannot be sure the survey questions were interpreted equally across participants; it is also likely that the framing of these survey questions has an impact on participant responses which may impact the generalizability of the findings. Despite these limitations, this study provides valuable insight to the design of haptic garments that provide compression and warmth, extending our understanding beyond typical vibration-based systems or pneumatic hug- based systems. This specific context of probing users’ expectations while in the design process, will help inform the development of haptic garments that mimic human touch and put researchers in a position to utilize similar SMA technology in the future. This is a first step into understanding users’ expected strategies and mental models used; future work will triangulate these findings with user studies involving physical haptic systems

User Expectations of Garment-Mediated Emotion Communication- Summary

In summary, we present an effort to characterize user expectations in selecting strategies of various intensities and patterns on different body locations for emotional communication through warm, compressive actuation, while delimiting/categorizing the range of mental models used during the expected ‘warm touch’ communication process. In particular, considering user’s mental models in relation to ‘warm touch’ parameters were particularly insightful and may have broader utility in understanding interaction paradigms for other haptic communication modalities. Broadly, we found 5 major mental frameworks, however their frequency of use is dependent on the emotion. Generally, 128 we observed that positive, prosocial emotions are more likely related to the use of social touch mental strategies while negative emotions are communicated as representations of body sensations. We also attempted to further understand possible effects of language on the communication of emotions (i.e., if there are differences in the ‘warm touch’ when the survey was framed as an emotional ‘communication’ vs. emotional ‘elicitation task). We found that for certain emotions like Fear and Happiness, there were significant differences between groups and saw that an egocentric versus allocentric perspective could have an impact in how participants deliver and understand the ‘warm touch’ haptic signals. It is hoped that these findings will inform the future research and development of affective haptic garments used for emotional communication.

129

Chapter 6

6. Compressive Haptics for Emotional Modulation

As we navigate through life, plenty of our experiences with the world and interactions with people around us are dependent on our ability to regulate emotions. Emotional regulation can be defined as the ways to which people experience, perceive, process, and react to emotional information, as well as the efforts to which people try influence when and what emotions they experience [183]– [185]. When done effectively, emotional regulation serves us well in maintaining both physical and mental balance, yet in many cases, proper self-regulation of emotions could be a struggle and has proved to be challenging for many; especially in the recent years where people’s changing lifestyles are resulting in rising levels of acute and chronic stress [184], [186]. Hence, building upon the work in the previous chapters, we will explore a high value application for compressive-based haptics; specifically, coupling on-body compressive haptics with mindfulness practices for long-term emotion regulation.

The idea to help people regulate their emotions is not new and there is an abundance of pharmacological and non-pharmacological interventions available, one of them being mindfulness

130 meditation. Mindfulness practices have been shown to be successful in reducing symptoms of anxiety/stress while providing a host of other physical benefits, believed to be due to modifying emotional regulation abilities in long-term practitioners [187]. Given that, many mobile technologies have been designed to support this practice [188]–[190], yet many of these interventions focus solely on software applications that leverage audio as a form of guidance/ support during the meditation process. Recent research in this area has shown promising potential for alternative interfaces of varying modalities for emotional regulation such as rhythmic light stimuli and tangible interfaces for emotion regulation purposes [30], [69], [189], [191], [192]. The work presented in this chapter leverages the known affective properties of touch [15], [112]–[114] to investigate whether use of warm, compressive forces provided by an SMA-driven soft robotic garment can provide a promising path as a non-pharmacological, non-invasive technological intervention for emotional regulation. In particular, we couple on-body compressive haptics with mindfulness practices that aim to provide a holistic meditation experience to evaluate its usefulness at encouraging users to engage in long-term emotional self-regulation practices 2.

Related Work

To understand the potential positive impact of wearable haptics for emotional regulation during meditation/mindfulness training, it is first necessary to briefly explore the theories and practices common to this domain. Meditation and meditation-inspired practices (often also referred to in Western settings as “mindful practices” or “mindfulness-based interventions”) can be described as a form of mental training through which practitioners develop control of their mental processes that eventually lead to the cultivation of greater emotional regulation abilities [185]. Such practices

2 This haptic-augmented mindfulness study was a collaboration with Justin Baker, Ph.D. student of UMN Human Factors and Ergonomics program and Crystal Compton, Ph.D. candidate of UMN Apparel Studies program. Justin has been an avid meditator and instructor with over 10 years of meditation experience and has provided valuable insights that helped drive the haptic garment design. Justin also participated in the user study as an instructor, guiding our participants through a short meditation practice, note-taking during the session, as well as providing copy-editing support of a subsequent conference publication. Crystal provided advice in garment patterning, base-garment construction help, and copy-editing support of the subsequent conference publication.

131 have been shown to provide considerable physical/mental health benefits, including reduction of blood pressure, heart rate, and stress levels [193], [194]. Diverse forms of this practice exist, but one key element across many forms of mindfulness practices involve the training of focused attention (FA) [185]; for a more detailed review on this topic, please refer to the work by Nash et al. [195] and Lutz et al. [196]. Contrary to the common misconception that mindfulness practitioners are simply engaged in relaxation, the practice actually also requires the ability to sustain and effectively allocate attentional resources; in other words, there has to be a balance between maintaining a certain degree of alertness while fostering a relaxed state of mind [185], [189]. In the context of emotion regulation, we contextualize mindfulness practices within an emotion regulation framework proposed by researchers in the field, known as the Process Model of Emotion Regulation. In this model, Gross (1998) proposes five phases of emotion regulation strategies, namely situation selection, situation modification, attentional deployment, cognitive change, and response modulation [184]; mindfulness meditation practices have been said to directly influence attentional deployment, resulting in the ability to exert better emotional control [187].

In a typical mindfulness session, FA entails maintaining focus on a single object. Whenever the mind wanders (construed as being absorbed in a tangential train of thought [197]) or when attention is drawn away from desired target of focus by unrelated thoughts, one is to gently bring attention back to the focal point without being discouraged (Figure 6-1) [197] [198]. As mind-wandering is a common human experience [198]; it may occur multiple times over the course of a session— resulting in a cyclical process and an integral part of the training of attention. FA is commonly interoceptive, directed at bodily sensations, typically the breath. Mindful breathing, taking slow, mindful breaths while focusing on accompanying sensations, is associated with decreased sympathetic nervous system activity [199], improved physiological states, and is even used in the treatment of anxiety/pain disorders [200], [201]. With continued practice, expert meditators tend to increase their capacity to be more aware of their internal states and better able to self-regulate [198], [202]. The neural correlates of mind-wandering appears to lie in the default mode network (DMN), a system made up of several brain regions including the posterior cingulate cortex and precuneus, lateral temporal cortex, dorsal and ventral medial prefrontal cortices, posterior inferior parietal regions, and hippocampal formation [198]. Evidence has shown that increased DMN activity is associated with negative mental health outcomes [202]. However, long-term practice is required to alter DMN connectivity [202], and FA is known to be a difficult process, especially for novices or individuals struggling with low-demand activities and may be one of the barriers to long-term

132 adoption of meditation practice and compliance [197], [198].

Attempt to Mind maintain wandering attentional occurs focus

Shifting of Awareness focus back of mind to breath wandering

Figure 6-1: A typical mindful awareness practice session—Mind wandering is a typical occurrence and happens in a cyclical, iterative process.

Given the understanding of the benefits that mindful awareness provides, there has been an influx supportive technologies, the most popular being mindfulness meditation apps that leverage audio output for guided meditation echoing [188], [189]. However, such apps usually only take advantage of one sensory channel and few have tried involving other sensory systems. Therefore, we take inspiration from Vidyarthi & Riecke’s Sonic Cradle in looking beyond mobile technologies while considering the holistic role one’s body plays in meditation practices [189]. As previously mentioned, in the process of FA, the body is typically used as the object of attention (i.e., the breath and their associated sensations), and thus presents a potentially impactful target for generating positive experiences or addressing attentional challenges.

In this work, we aim to involve the body to cultivate a positive experience of mindful meditation without complex effort demanded by traditional meditative practices. In particular, we look towards the sense of touch; it is known that touch is an immensely rich communication channel [31] and is natural as it is common to human behaviors [190]. From a non-technological perspective, there are forms of meditative practice which involve the sense of touch, but it's rarely considered (primarily at least) as a way to assist FA; an example would be the use of beads during the meditation practice [203]—however this is a physical task and has distinctions with passively applied haptics. To our knowledge, applied, passive systematic haptic stimulation has not been used in traditional meditation practices likely because meditation practices are typically done alone. However, given the positive effects the sense of touch may provide, haptically-mediated technological interventions could prove to be an interesting avenue to explore. Given that, some researchers have designed technologies that has been showed to be successful in lowering anxiety levels, demonstrating how 133 haptics can be used to modulate affect. Costa et al.’s EmotionCheck wrist-based wearable system utilized vibrations to regulate a user’s anxiety by providing users a false feedback of slower paced heart rates [69]. In contrast, Papdopoulou’s Affective Sleeve system builds upon that idea but instead utilizes warmth and pressure applied on the forearm that that is intended to better reflect flow induced rhythmic haptic sensations closer to human touch sensations [30].

Building upon these past works, the goal is to investigate whether we can leverage compression and warmth stimuli, replicating the beneficial effects of human touch, in cultivating a positive experience in mindful meditation. However, unlike other wearable solutions, we shift the focus away from the wrist/arm, towards other larger, distributed areas that could reflect common human behaviors (e.g., hug, shoulder massage). In contrast to some haptic interventions that employ vibration feedback, this work uses compression and warmth sensations as they have been shown to be successful in promoting the relief of muscle tension, encourage stress reduction, relaxation in the state of alertness, reduction of mental anxiety, serving as a form of emotional ‘grounding’ and provide an overall improved feeling of well-being [9], [22], [117], [118], [129].

With that, we designed a haptic garment that applies compression and warmth on the shoulders rhythmically to provide a naturalistic and holistic mindfulness meditation process. It is believed there is potential for this form of technology (garment capable of delivering compressive and warmth sensations) to provide the stated benefits since past evaluations on similar garments has participants associating this form of garment use with emotional modulation scenarios (Table 4-1). Specifically, we will investigate if the garment, and the bodily sensations it stimulates, will act as a cue for grounding attention within the body (and away from thoughts). Supported by the work of Levinson et al. [197], who found that low-demand tasks like meditation may lead to mind- wandering, we also aim to ground attention. We hypothesize that the cyclic haptic stimuli, while increasing bodily sensations, may attract more attentional resources to the relevant target— the body, and leave fewer resources available for wandering. In this chapter, we investigated if the use of wearable compressive haptics can aid novices with FA during mindful meditation.

Compression-Actuated Haptic Garment Design

The haptic garment design leverages selected guidelines put forth by Adams et al. for designing mindless computing technologies and employs the novel soft robotic technology-integrated garment architectures developed and described in this dissertation [204]—these guidelines were 134 selected because ideally this technology would sit between the edge of consciousness during the meditative practice, balancing between a state of alertness and relaxation, as well as the broader goals what this technology could be in the future—a form of persuasive technology that could subtly influence the behavior of users without requiring concerted awareness (a concept which will be discussed later in the chapter). Hence, the technology was designed with the following guidelines in mind: (1) it should be reflexive and straightforward (i.e., engaging the automatic mind to trigger behavior changes without conscious thought), (2) it will be more effective if somehow embedded into people’s daily routine, (3) cues acting in the periphery should not be very different from what user would normally experience in daily life. With those guidelines and support from past literature in mind, a garment system was designed to provide rhythmic haptic action with compression and warmth, replicating the known pleasant aspects of human touch [9], [205], and the system was designed with compression-based actuators integrated into a wearable garment system, using materials found in everyday clothing. The garment and subsequent study design were also constructed based on the feedback from two expert meditation practitioners with over a decade of meditation practice and instruction.

Figure 6-2: Men (Left) and Women's (Right) compression garment prototype with demonstrated garment (comfort/under) layer and SMA compression actuators located on the shoulder regions.

The prototype compression garments (Figure 6-2) utilized coiled NiTi SMAs to create dynamic compression, similar to those in Chapter 3 and 4. The SMAs apply compression through coil contraction when heat (or current) is applied and the forces they provide scale with applied current, affording system controllability [160]. SMAs provide both compression and warmth stimuli, induced by single material actuation. The SMAs were housed within braided sheaths (Techflex ¼”) for electrical isolation and to support cyclic resetting of SMAs (Figure 6-3)—for detailed

135 mechanical operation, refer to Chapter 3, section 3.4.

Six actuators (16.5cm) were attached on each shoulder vertically using snap connectors, ½” apart. Each SMA received ~0.6A of current, each providing a linear force up to 7.1N based on mechanical tests (Figure 3-16). The shoulders were selected as the regions of haptic presentation for a variety of reasons, (1) from past studies, the shoulders were found to be more sensitive to the perception of compressive sensations than the torso (Chapter 4) [168], (2) the torso was avoided as to not interfere with focused breathing and (3) compression on the shoulders are similar to common human behaviors and is commonly associated with calming actions (Chapter 5). The actuators were programmed to actuate for 5 seconds and unpowered to relax for 12 seconds—this was selected based on pilot tests on the garment to maximize perceptibility of the stimulus. The garment was remotely controlled through Bluetooth (MOSFET-driven), using Processing user interface to control stimuli presentation through a computer.

The garments were designed for both male and female, size medium, and consisted of 2 layers (Figure 6-2): (1) an outer actuation layer with 2-way stretch Neoprene knit fabric (non-stretch oriented in the lengthwise direction to which the SMAs connect), with Velcro© closure for adjustability and an elasticized silicone-lined hem for garment anchoring. (2) an under-comfort layer (athletic crew neck long-sleeved knit shirt 90% polyester, 10% spandex) with an insulating Neoprene shoulder pad for heat generated by SMAs. The SMA-integrated shoulder area surface directly interfaced with the skin was 34oC after 8 mins of actuation.

A textile force sensitive sensor based on the design by [206] (i.e., two conductive materials sandwiching Eeyonyx material) was integrated into the top of shoulders to observe force changes 136 with compression, showing users were exposed to averaged pressures of approximately 15- 24mmHg (the sensor manufacturing, calibration, and resulting pressure estimations calculations are described in Appendix H). Two breath rate bands were worn in-between the comfort and actuation layers, one below the chest/above the diaphragm, and one on waist—two sensors were used for redundancy to accommodate for both ‘chest breathers’ or ‘diaphragmatic/belly breathers’ (Figure 6-4). The sensors were manufactured according to [206]; they were coverstitched (length= 5”, with one trace replaced with silver-coated nylon 4-ply 235/34 dtex conductive thread) stitched on 84% polyester/16% spandex knit, and both ends connected to non-stretch cotton canvas fabric with Velcro attachments for accommodating varying participant sizes. The coverstitched sensor is responsive to stretch, given torso/abdominal expansion or volumetric changes as the participant breathes. Varying resistance can be captured due to changes in configurations of the circuit in the looped structure of the stitched conductor in response to stretch [206], [207]. The soft, conformable form factor was the reason the breath sensor was constructed in-house instead of using commercially available devices to preserve wearer comfort and avoid potential distractions from rigid components during the meditation process.

Data for both the breath and pressure sensors were collected using National Instruments USB-6001 data-acquisition system, used to supply a constant 5V power source as well as to record analog voltage from each addressable sensor trace. A 5.6 kOhm resistor was used to create voltage-dividers and data from the DAQ were sampled at 5 Hz and recorded using MATLAB SimuLink software.

137

Study Design and Data Collection

The focus of this study is on evaluating the role of the haptic garment in augmenting/mediating the mindful awareness experience, as well as the efficacy and potential use of such wearable technologies in mindful meditation practice for novice meditators. A within-subjects study design was employed (n=10, 4M/6F, age 19-26, mean=21; participants were recruited from the university through posters and word-of-mouth) and included only novices (<3-month meditation experience; participants were screened prior to arriving for the study and before the consent process). The study consisted of 5 phases (Figure 6-5). Subjects were first fitted with a biometric-collection wrist device and completed baseline surveys. Then they were guided through a 2-minute practice of Samatha meditation, a cognitively directed method in which the participant’s eyes are closed and attention is concentrated on the breath [195]. This guided task was performed by an experienced meditation instructor (over 10 years’ experience).

During practice, subjects were also instructed on the way of tracking their attention, by pressing a button on the wrist device whenever the mind wanders. The 2 meditation tasks lasted 8 minutes each; the control (without garment) and intervention (with garment) presentation orders were randomized. As a within-subjects study design, all 10 subjects experienced both with- and without- garment conditions. The study was conducted in a 10 by 10 ft. closed room (private office space equipped with central ventilation, well-lit, and isolated from most external noise) and subjects were seated during the tasks. Subjects completed questionnaires after each 8-minute meditation task about their experience and participated in a qualitative interview at the end of the study. This study was designed with guidance and input from 2 meditation experts with over 10 years of teaching experience.

Establishing baseline conditions

2-minute guided meditation practice

Mindful meditation task 1

Mindful meditation task 2

Qualitative interview

Objective Measures

Biometric data was collected using an Empatica E4 wrist device; in this study, we use only electrodermal activity (EDA) and heart rate data. For each participant, HR data was averaged across each meditation period. EDA was normalized from 0 (min) to 1 (max) to eliminate differences in individual ranges. Using Ledalabs software’s continuous decomposition analysis, Skin Conductance Responses (SCR) of each participant were calculated for each meditation period (using a threshold of 0.05μS) and averaged across participants. The Empatica E4 was also used to track instances of mind wandering; whereby participants pressed a button on the device as an acknowledgement of mind wandering before returning to their breath. This use of a button was intended to replicate a common practice in certain mindfulness meditation practices (Zazen meditation) whereby practitioners follow a specific ritual upon detection of loss of attention before returning to their object of focus such as the breath, and it is a method that has been used in other research studies as a means of tracking attention [198], [208]. The number of button presses in each meditation task was calculated and averaged across participants.

Figure 6-6: A representative graph (P1) collected from BR sensor. Note the prominent peaks corresponding to each breath.

To track breathing rates, a low-profile, soft stretch sensor was developed (Figure 6-4) [206]. Two sensors at different locations (chest and abdomen) were used (Fig. 8) to capture diverse breathing styles; but only data from the sensor with larger amplitude changes were analyzed. The sensors were validated against manual thorax rise/fall counts by an observer. There was an average discrepancy of 1.58 ±1.02 breaths over 8-minutes; this difference likely stems from differences in start/stop periods between the observer and sensor. The peaks recorded were extracted over 8- minutes (Figure 6-6) and the number of breaths per minute for each participant were computed. 139

Subjective Measures and Qualitative Interview

The subjective questionnaire consisted of (1) State-Trait Anxiety Inventory (STAI) to measure changes in anxiety [209], (2) Flow State Scale (FSS) [210] measuring enjoyment, concentration, perception of time, and (3) experiences with 9 semantic differential scales (SDS). FSS items are 5- level Likert scale (completely disagree to completely agree) with levels of agreement mapped to 1- 5. The 5-level SDS featured differential labels anchored at the extremes. A qualitative interview was also performed at the end of the study and participants were encouraged to share their thoughts. Questions were posed for self-assessment of focused attention, comparing experiences with and without garment (i.e., if they felt they performed better with the garment or otherwise), and for time perceptions. We also probed their insights on the haptics experienced, meditation preference (either they preferred the experience with or without garment), likes/dislikes in regard to the experience, and thoughts on the system’s purpose/future use. The interview data was sorted using affinity mapping strategies to cluster ideas and extract themes.

Results and Discussion

Overall Compression Experience

The compression was well-received by most participants; the sensations were said to be perceivable but was not of overwhelming intensity such that it was ‘almost forgotten [when meditating]’ (n=3). 4 participants commented on the unique, participatory method of meditation; 2 subjects even remarked that the meditation experience felt lacking without the haptic sensations provided. However, many stated that the initial compression was strange or alarming (n=7); they eventually acclimated, after which the comfort was akin to regular clothing (n=3). Specifically, the rhythmical pulsing was said to be similar to or reminiscent of slow breaths (n=3). While the garment reception was mostly positive, some (n=3) did not care for compressions provided—consistent with past studies where some individuals are averse to on-body haptics [137]; they felt that the compression haptics were not suited for meditation, for reasons including limited gut expansion for breathing (n=1), unpredictability of the haptics/non-synchrony with their breaths, which in turn, was agitating (n=2), or resulted in an induced sense of claustrophobia (n=1). We observed that compression perhaps is not universally enjoyed; in particular for this application, participants’ preferences for compression sensations were bifurcated into affinity and aversion groups.

140

Interestingly, unlike the previous garment evaluations using similar actuation technologies (Chapters 3 & 4) the garment warmth was very well-tolerated by almost all users; the garment warmth was reported to be very well-liked (9 out of 10 subjects). The more tolerable thermal sensations are likely a direct design impact of employing pulsing sensations (that inadvertently reduced power usage, and hence heat generated) and changing insulation material utilization. The warmth presented by the garment was said to be of the right intensity (n=9) and it warms up not only the shoulder areas but envelops the whole body. The sensations were described to be similar to actions provided for physical relaxation (massage, swaddling, or the use of warm, weighted blankets), comfort (akin to a hug, shoulder rub, or being close to a warm fire), or Autonomous Sensory Meridian Response (ASMR)-like experiences, i.e., warm, tingling sensations that are said to be deeply calming/relaxing (getting into a hot tub, getting hair done) (Figure 6-7). With these participant-reported experiential effects in mind, we delve deeper into some other themes provided.

Figure 6-7: Participant responses in comparing sensations provided by the SMA-based compression garment to everyday activities.

Direct Physical Benefits

Many participants commented that with the warmth and compression sensations provided, they could feel their muscles physically relaxing (n=5) and the garment’s pulsing motions helped them get into a rhythm of breathing (n=5). With that, we look into the physiological data collected. Since we observed reported differences in preferences (i.e., some participants were averse to the compression sensations during meditation practice), we present and organize the following quantitative measure results based on the bifurcated self-reported preferences of the compression

141 intervention: ‘liking’ (solid, green) and ‘dislike’ (shaded, red). All bar graphs were presented as differences of control minus intervention (i.e., a positive number means intervention had a lower measure than control, vice versa) and the x-axis participants were rank ordered from highest to lowest measure scores. Figure 6-8 to Figure 6-11 presents results extracted from the breath rate sensor and Empatica E4 device (HR, EDA, self-report loss of focus).

First, we investigate the effects of the compression garment on participants’ breath rates. From Figure 6-8, for n=5, the intervention was successful in lowering their breath rate (positive values), n=4 had an inverse effect, and n=1 had no change. Interestingly, these results aligned well with subjects’ preferences; P3, P4, P5 and P8 were in the ‘dislike’ group and we can see from Figure 6-8 that they had equal or higher breath rate with the garment than without. From the Empatica E4, we also gathered heart rate and EDA data. First, in terms of heart rate, all meditation tasks (control and intervention) decreased HR compared to baseline; the grouped average baseline HR (80.8±10.7 bpm) is higher than both control (73.6±14.2 bpm) and intervention (73.3±14.6 bpm). Figure 6-9 presents the differences in heart rates between control and intervention. Somewhat consistent with preferences, 3 out of 4 of those in ‘dislike’ group (P5, P8, P4) all had higher heart rates with the garment; with those of the highest heart rates belonging to P4 and P5 from the ‘dislike’ group. For EDA, SCR counts are presented in Figure 6-10. Here, we see a trend where those in the ‘dislike’ group had lower SCR counts for the intervention than control. This decrease in arousal compared to the control may be a reflection of their subjective reports of annoyance/distraction towards the stimuli. However, since EDA only detects arousal, conclusions based on the quantitative measures alone cannot be drawn.

Figure 6-8: Differences in breath rates between Figure 6-9: Differences in heart rates between control and intervention; results were rank ordered control and intervention; results were rank ordered for the x-axis. Participants P3, P4, P5, P8 did not for the x-axis. Participants P3, P4, P5, P8 did not prefer the haptic garment. prefer the haptic garment.

142

Figure 6-10: Differences in the number of SCRs Figure 6-11: Differences in the number of instances between control and intervention; results were rank of loss of focus between control and intervention; ordered for the x-axis. Participants P3, P4, P5, P8 results were rank ordered for the x-axis. Participants did not prefer the haptic garment. P3, P4, P5, P8 did not prefer the haptic garment.

Emotional Effects

From participant responses, it was found that the compression and warmth stimuli provided by the garment lead to subjective emotional changes. Many participants said the sensations provided by the garment were calming/relaxing (n=6) or comforting/secure (n=5). From STAI results for ‘liking’ group, the STAI scores decreased by an average of 4.17 points for the control group and 8.83 for the intervention group compared to baseline. In contrast, for ‘dislike’ group, STAI scores decreased by 4.50 points for the control group and only 2.25 points for the intervention group compared to baseline. The improvement aligns with expectations for meditation; liking the garment enhanced the anxiety-reducing effect, while disliking it partially canceled the effect.

Comparing that to the SDS items (Figure 6-12), overall, it showed that the garment intervention on average, provided more feelings of being active, positive, and warm than control (no garment). There were no observable differences between control and intervention in terms of pleasantness, comfort, relaxation, and dominance, which may mean that the garment and its haptics were not adversely affecting subjects’ comfort levels.

Cognitive Effects

Perhaps more interesting are the cognitive impacts as a result of the stimulation provided by the garment. One major theme that emerged was concentration and focus. From the attention tracking task (Figure 6-11), n=5 had improved attentional focus (positive values-green) and n=4 had worse attentional focus (negative values-red), and n=1 had the same results across both tasks. From 143 interviews, P3, P5, P8 from ‘dislike’ group felt the garment worsened their attention (stimuli drew attention away from breath, which made the mind wander more (n=3)), consistent with Figure 6-11 measures displaying equal/worse performance for intervention compared to control. However, while P4 was from ‘disliking’ group, when asked about garment distraction in the qualitative interview, she did not think it was distracting except for the first actuation, which was consistent with her attention results. For the n=7 subjects that thought garment improved their focus, the haptics were said to reduce distractions from external stimuli (n=4), resulting in fewer thoughts during meditation (n=3). One participant even pointed out that even when mind-wandering happens, it felt as though it was for a shorter amount of time. As reflected in the FSS (Table 6-1) item Q5 and Q23, for the ‘liking’ group, attention was said to improve with the garment intervention. In contrast, the FSS’ self-reported attention ratings declined with garment intervention for the ‘disliking’ group.

The garment was said to act as a focal point (n=4) to help those who have too much on their minds have a more regulated meditation process. Some thought that with the garment, the meditation seemed to have required less conscious effort to keep attention on breath (n=3), ‘like I wasn’t trying at all’ (n=2). Evidenced by FSS item Q14 (Table 6-1), for the ‘liking’ group, we see that the garment intervention resulted in better self-reported performance in regards to effort required to keep the mind on task compared to without the garment; while the disliking group reported no average changes. One interesting account was provided by a participant (P6), who said that he wanted to try the practice but had to abort past meditation attempts because he would get caught in

144 a negative feedback loop of thoughts and would not be able to continue meditating. He compared the garment haptics as being a ‘…nice distraction that pulls me out of my thoughts with a focal point’, and a technique to help those in a panic attack, ‘… like having only five things to focus on in a room’. Along those lines, some also commented on an enhanced body awareness with the intervention. The haptics were said to draw focus to the body (n=2) and created more ‘centered’ or ‘grounded’ feelings (n=3). In particular, P7 commented that, ‘… usually it is difficult getting the mind to isolate itself. Typically, we think of the mind needing to take control of the body. With the garment, it feels like the body is not my enemy and it's actually helping me’.

Table 6-1: Selected Flow State Scale (FSS) items

Q. Avg. 'Liking' Group Avg. 'Disliking' Group Item ID Control Int. Control Int.

5 My attention was focused entirely on what I was doing. 2.8 ± 1.3 4.0 ± 0.6 3.8 ± 1.3 2.3 ± 1.3 8 Time seemed to have sped up. 2.0 ± 1.1 2.7 ± 1.4 2.3 ± 1.3 3.8 ± 0.5

9 I really enjoyed the experience. 3.7 ± 0.5 4.7 ± 0.5 4.0 ± 0.8 3.5 ± 1.3

14 It was no effort to keep my mind on task. 2.0 ± 0.6 3.2 ± 1.0 3.0 ± 1.4 3.0 ± 1.8

27 The experience left me feeling great. 3.8 ± 0.8 4.3 ± 0.8 4.0 ± 0.8 3.5 ± 1.0

23 I had total concentration. 2.2 ± 0.8 3.0 ± 1.1 3.8 ± 1.3 2.5 ± 1.7

36 I found the experience extremely rewarding. 3.3 ± 0.8 4.0 ± 1.1 4.0 ± 0.8 3.3 ± 1.0

Another theme was the garment’s role in altering the perception of time. From FSS (Table 6-1) Q8, time was perceived to pass faster with the garment for both ‘liking’ and ‘dislike’ groups. This was also reflected in the interview, where participants commented that with the garment, they weren’t worried about time (n=3) and that time passed much faster with the garment than without (n=4). With fewer thoughts about time, they felt more immersed in a meditative state and could have continued for a longer session (n=3). Such reports of more immersive/pleasant experiences may reflect some similar elements to the reduction of DMN activity that is associated with skilled attentional focus [211].

Compression-Augmented Mindfulness Study Summary

6.5.1. Discussion and Limitations

The designed haptic garment received generally positive feedback; the warmth and compressions were largely welcomed, though as in previous studies we noted some individual differences in how

145 compression is received [168]. While many report positive opinions of compression as a mode of haptic stimulation, some individuals generally (and fundamentally) dislike the sensation of compression, and generally responded negatively to the study’s intervention. This concept of individual preferences and responses to stimuli is an important consideration in designing and deploying these types of technologies; in fact, some researchers who investigated the use of a wearable device (MoodWings) designed to calm users resulted in the contrary and they acknowledged the existence of a fine line between a stress intervention and a stressor if not employed in the right context [212]. Overall however, we found evidence that haptic assistance shows preliminary value as a support or intervention tool for improving the meditation experience and helping individuals with FA, particularly for those who report affinity for compression as a form of on-body stimulation.

Revisiting the initial goal to help novices in reducing mind-wandering, results show that garment haptics has the potential in helping hone self-perceived focus (70% reported less distractions). However, no conclusions can be drawn since tracked attention demonstrated only 50% of subjects performed better with intervention. In other words, the quantitative data was more ambiguous, but the majority of participants reported improved sense of focus with the garment. This discrepancy between self-reported perception and captured mind-wandering instances should be further examined, but it does bring up the question if the task of button presses itself may have provided some form of distraction in the first place, or perhaps the subjective perception of an improved performance could serve as a sufficient motivator to encourage users to participate in mindful meditation activities long-term.

Also promising is the fact that the feedback on enhanced body awareness and immersiveness appears in line with our initial theory that increasing bodily sensations may draw more attentional resources and reduce mind-wandering. We believe the garment may have facilitated a more holistic, immersive experience, involving both mind and body, encouraging novices to leverage compression and warmth on the body to guide attention back to a focal point (i.e., having body help regulate the mind), effectively restoring one’s inward attention. However, the positive effects were only observed for those who liked the haptics. The feedback did not rule out that different haptic stimuli, e.g., different intensities/patterns/locations optimized for this specific application (e.g., incorporation of a feedback loop that provides cyclic compressive forces that matches a user’s breathing rate), could improve the experience for users in both the affinity and aversion groups,

146 and should be an emphasis for future investigations in this space.

Many also described the haptic garment as a ‘calming’ or ‘secure’ experience, but currently, what that means exactly and how that affects mindfulness needs additional study. Note that in our introduction of the mindfulness practice, we mentioned that there needs to be a balance between maintained attention and a relaxed state of mind, a natural concern was that this balance was tipped with the introduction of the haptic garment intervention. From our study, although ‘calming’ connotations were implied in participants’ feedback, from Figure 6-12 SDS results, we suspect that users are not overly relaxed to the point of dullness since they reported feeling more activated with the garment than control—implying a sense of balance was achieved for at least some participants. More work has to be done to understand the garment tradeoffs in allowing users to pursue a balance between the two components. Along those lines, it would also be important dissociate how each of the two haptics (compression/warmth) contributed to different aspects of the experience. Perhaps, the addition of additional objective biometrics such as electromyography (EMG) (measuring muscle activation), photoplethysmogram (PPG) (measuring blood volume changes), or electroencephalography (EEG) (measuring electrical activity of the brain) could provide further insights into some of the questions posed above.

Finally, there are additional limitations to the current study that we hope to remedy with future work: first being the small number of participants recruited for this investigation—especially when we realized the need to divide participants based on preference for the compressive sensations, which did not provide us with enough power for robust statistical analyses (and hence was not presented), or the ability to draw major conclusions in regard to the effectiveness of the garment intervention. Secondly, while some biometrics were presented, it was not intended to drive conclusions because the mindfulness experience varies for each participant, and physiological benefits can only be seen with long-term practice. It is hoped that by improving the overall experience and encouraging repetitive behaviors, we can build upon this work and observe long- term benefits. Hence, longitudinal studies should be conducted to understand the true benefits provided by this compression-augmented meditation practice. Further, this study did not establish a true baseline (baseline was captured when users were completing surveys)—ideally, the baseline measures will be captured for a longer period of time, without having participants performing additional task. With further research, we hope that the haptic system will be able to inspire new practitioners through an improved meditation experience and long-term adoption of the practice.

147

6.5.2. Broader Implications for Persuasive Technology in Emotion Regulation

The SMA-based haptic garments, capable of applying distributed on-body warm, compressive forces, were used as a way to augment mindfulness meditation practices. It is hoped that the garment can help cultivate a positive experience of mindful meditation without complex effort, help novices engage in these practices long-term, which would eventually result in emotional regulation benefits. Emotion regulation strategies are sometimes discussed in terms of whether they are ‘explicit’ or ‘implicit’, the former being strategies that require conscious effort or awareness by individuals that they are regulating their emotions, while the latter involve emotional regulation automatically, without conscious supervision or intention [213], [214]. There has been debate on whether mindfulness meditation is a form of implicit or explicit process and the practice is currently believed to be a mix of these strategies [215]. FA is a reflective process that is typically not performed concurrently with other tasks due to the need for maintained attention. Recently, researchers are investigating alternatives that could move the practice towards the implicit end of the spectrum, through technologies that can subtly change attitudes/behaviors of users without needing conscious effort—which could include forms of persuasive technologies [106], [204].

Biofeedback is a commonly integrated persuasive technology strategy that has shown considerable success for emotion regulation; an example would be the work by Costa et al. that has shown it is possible to regulate a user’s anxiety levels by providing false feedback of a slow heart rate (that resembles a pulse) [69]. In our research efforts presented in the sections above, the designed SMA- based haptic garments (in their current form) are not a form of persuasive computing, but they have tremendous potential to help transform attention regulation from an explicit to a more implicit process. There are several reasons as to why we believe this technology is in a unique position for success:

(1) The use of compressive forces is successful and prevalent in the field of occupational therapy (DTP) as a method to induce calming effects, serving as a form of ‘emotional grounding’ [9], [22], [129]. This demonstrates a highly plausible link between compression stimulation and attitude and/or behavioral changes.

(2) The wearable garment form factor that interfaces directly with the user can provide intimate haptic feedback without interrupting the user in a socially acceptable way, and is readily embedded 148 into users’ daily routines (i.e., clothing). This aligns well with the recommendations provided by Adams et al. in designing mindless computing technologies that fades into the background of the user’s everyday experiences [204].

(3) The garment form factor also serves as an excellent platform for the integration and coupling of other forms of sensors. For instance, sensors that are capable of autonomously detecting a users’ physiological state or contextual cues and adapt the compressive sensations accordingly.

(4) Warm, compressive sensations are known to be more pleasant than other forms of tactile feedback; this is in contrast to other forms of biofeedback technologies currently available that utilize vibrotactile actuation that displays a user’s false/slower heart rate to regulate anxiety [69].

In this context, we can imagine a garment applying subtle, warm compressions on the shoulders based on a user’s heart/breathing rate to help regulate emotions for everyday use. According to various contextual cues, the compression parameters can also vary. For instance, depending on whether it is an acute intervention, or an activity targeted at long-term emotion regulation goals, the haptic compression pattern and intensity can also be varied accordingly.

Further, this current study investigates the use of the haptic garment in a ‘neutral to low arousal context’ (i.e., through a mindful meditation study); to fully appreciate the breadth of possibilities this technology could provide, the next step is to understand the effects when this technology is used in a ‘neutral to high arousal situation’, i.e., when users are acutely experiencing a stressful situation. One way this could be done is to deploy the designed garment in a case where users are experiencing an introduced stressor. With that, the objective and subjective effects provided with the haptic garment can be systematically studied and compared to instances without a garment intervention—a proposed study is presented in Appendix I. Given the investigated effects derived from the user study presented in this chapter, as well as proposed benefits highlighted in this section, we believe that the developed SMA-based haptic technologies could provide a wide variety of opportunities in the use of wearable persuasive technologies for emotion regulation.

149

Chapter 7

7. Conclusion

This dissertation investigates how people experience compression stimuli on the body given varying compression inputs, use-cases and contexts—using a novel body-scale garment platform integrated with soft robotic actuation technologies to deliver compression stimuli. This research demonstrates the presentation of compressive forces using active material actuation technologies as a proxy for human touch, enables novel wearable interfaces and interaction possibilities. In this chapter, we summarize key findings from our research and discuss broader impacts of this work in the field of human factors, haptics, and wearable technology.

Summary of Key Findings

We contextualize key findings for each chapter within a human-machine interface framework (Figure 7-1).

150

Figure 7-1: Human-machine system. Image obtained from [216]

In Chapter 3, we presented an end-to-end design process for wearable compression garments integrated with soft robotic actuation technologies that are capable of delivering variable compression inputs on distributed, upper body areas. Here, our investigation centers on the ‘machine’ end of the human-machine system, in developing technologies that will enable the study of questions relating to the human experience of compression. Keeping the core idea that the humans are part of the system and therefore should not be ignored, we take a human-centered design approach to the system design—for each major design iteration, user studies were performed to gather user feedback. Starting with rapid prototyping (using readily available materials) that resulted in a low-fidelity prototype, we were able to gather immediate insights from users that were crucial in developing following technology-integrated compression garments. In particular, we were privy to the importance of incorporating design strategies that emphasizes the need for accommodating movement in body locations such as the abdomen and shoulder joint, the importance of material selection, gender-separated designs to accommodate for anatomical variances, and placement of actuators on different body locations that resulted in different compression sensations and vectors, just to name a few. Given a better understanding of the design requirements needed from user feedback, we were able to further investigate the feasibility of using active material (SMAs) as an actuation scheme and continually iterate on the design that finally resulted in a high-fidelity prototype that we could use as a research tool to study the effects of compression on the body. Some key highlights of the high-fidelity prototype include the ability to produce computer-mediated, dynamic compression via remote Bluetooth control, with varying compression inputs of (location, duration, intensity, pattern) that emphasizes soft form factors for enhanced wearability. The research efforts undertaken resulted in several contributions to the realm of soft robotics-integrated wearable garments.

151

Given the advances achieved through the SMA-based compression garment system, in Chapter 4, we presented our efforts to assess the subjective, experiential effects of computer-mediated compression, given varying compression stimuli parameters presented on various upper body locations. Specifically, we were interested in the ‘human-interaction’ components of the human- machine system framework. In this study, we focused on the resulting physical (i.e., perception of compression) and perceived cognitive impacts (i.e., comfort and preferences) when compression of varying intensity, duration, and pattern were presented on different upper body locations. From a user study with 17 participants, we observed several major themes including how compression vectors influence user’s perception and understanding of a stimulus, the need for customizability due to varying individual preferences, as well as confounds that exist with the design and use of compression-based technologies including anatomical variances that influence sizing and fit, as well as the relationship between contextual elements and user perception or interpretation of the compression stimulus. The key contribution from this investigation include an increased understanding of the use of compressive haptics as perceptual/experiential interaction stimuli. In the context of human factors, this research provides an added understanding of touch perception provided by mediated SMA-based compressive technologies.

Understanding of the physical and subjective cognitive impacts of on-body compression stimulation, we were interested in the potential of using compression as a mode of communication. Specifically, in Chapter 5, we investigated user expectations in using varying warm compressive actuation forces (as provided by SMA-driven compression garments) to communicate 7 distinct emotions. We were not only interested in the strategies people use compression-based interface to communicate these emotions, but also the cognitive frameworks employed. Through two online surveys, we noted 5 major categories of mental models used by participants to make sense of this form of garment-mediated emotional communication, including the representation of body sensations, replication of typical social touch strategies, metaphorical representation of emotions, symbolic representation of physical actions, and mimicry of objects or tasks. Given the gathered frequency of use of each of these mental frameworks, we were able to map the design space for future haptic communication applications utilizing soft robotic-actuated compression garment technologies. The key contribution of this research work is the identification of the challenges and opportunities of SMA-driven compressive-based haptics as a communication tool—extending our understanding of compressive haptics beyond just touch perception, but also towards possible cognitive frameworks relating to haptic strategies during the communication of emotions.

152

In Chapter 6, we took a step back and studied the human-machine system as a whole, investigating how the compression garment system can be applied to and embedded within a larger use context. Specifically, leveraging the known positive effects provided by warm, compressive on-body stimulation, we demonstrate the use of the designed haptic technologies in an emotion modulation application through a compression-augmented meditation task. The majority of novice meditators using the haptic garment reported an improved sense of focus, enhanced body awareness and immersiveness (sped up perception of time), and improved measured biometrics including (reduced heart rate and breath rate). However, these positive effects were only observed for those who liked the haptics—some individuals generally (and fundamentally) dislike the sensation of compression. Overall, we observed tremendous potential in using SMA-driven warm, compressive on-body stimulation to positively augment meditation and discussed implications for emotion regulation applications. The contributions derived from this work lay a foundation for future haptic-based persuasive technology applications.

Limitations and Future Directions

We acknowledge that this research has some limitations that warrant discussion and future investigations. While the these have been introduced in the relevant chapters, we list some broader themes and overall directions to pursue for future work.

First is regarding the choice of SMAs as the actuation mechanism for the application of compression. SMAs in the introduced coiled configurations are capable of delivering compressive forces onto the body with an applied current while maintaining small and soft form factors. However, there are some inherent drawbacks in material properties that will affect the systems and applications in which they are employed. SMAs have low frequency response which makes them unsuitable for systems requiring highly dynamic responses. Further, since the shape change capabilities of SMAs are thermal dependent, it means that warmth will be generated during the actuation process—and hence heat accumulation over time. In the application scenarios described in this dissertation, it could add another layer to an interaction when viewed as a proxy for human touch—but the added, accumulated thermal stimulus may be detrimental in other forms of application and/or if left uncontrolled. For instance, (1) when the thermal stimulus becomes too hot/warm which detracts from the compression experience, or (2) when disassociation between thermal and warmth stimuli is necessary).

153

Further, the nonlinear properties of SMAs such as dependence on temperature and stress, as well as phase transformations makes mechanisms required to precisely control SMAs particularly complex [150], [151]. The relationships between temperature, forces, and power have been modeled but is typically done so without a human in the system [123]; since the human body is also a dynamic system (both in terms of changing surface pressures as a person moves and/or as a thermal sink), these models may be limited as a way of precisely predicting applied stimuli for closed-loop controls. Another way this could be circumvented is the use of sensors to measure pressure and temperature stimuli real-time for each user—this remains a limitation of the current research since our implementation does not involve a sensing system. Advances in this area could allow for more personalized and customizable experiences. However, pressures on the body are difficult to measure. The human body is a soft, contoured system and there is currently no commercially available device that can accurately measure absolute pressures on a human body surface due to limitations in the calibration process. Hence, more research has to be done to implement control mechanisms to optimize and control the deliverance of warm compressive forces on a human body, which could include the implementation of sensors and algorithms to detect, predict and adapt to user and context-driven needs.

SMAs are also notoriously power hungry; to actuate SMAs within 5 seconds, each actuator requires ~0.3A of current (for quicker actuation times, more power is required, which will incur higher compression sensations and warmth). Future work should investigate design strategies for optimizing the relationship between power, temperature, and forces generated by SMA actuators depending on specific applications. Some potential avenues of research could include investigating different material combinations of the SMA actuators to optimize performance (the NiTi Flexinol wires used to form SMA coiled actuators in this dissertation were chosen due to their commercial availability). Varying material combinations coupled with different shape-setting procedures could alter the material’s transformation temperature windows, changing the actuator’s performance such as generated force, power consumed, and the amount of waste heat generated (preliminary investigations [162] [174] has shown promise of this method in reducing power consumption, and in some cases, even harnessing heat generated by the body for SMA actuation. This could mean there is a possibility of requiring low- to no external power source for SMA activation). However, these investigations are still in preliminary stages and should be further pursued. Expanding upon the idea that different SMA materials could be good alternatives, we should also point out that SMAs are also not the only way of inducing actuation-- many other forms of soft robotic technology

154 also exist; so long as the garment form factor and actuators can be developed in comparable capacities, the findings here could potentially be translated or the design of technologies and capabilities be expanded.

Another alternative is to involve mechanical/electrical design strategies instead of altering the SMA actuator’s material properties. For instance, the use of mechanical designs that can capture and maintain compression created during SMA activation (e.g., a latching mechanism that will mechnically ‘lock’ the compression, enabling supplied power to the SMAs to be turned off), or electrical means (e.g., pulse width modulation) to reduce power consumption. SMAs are theoretically very suited for these types of control mechanisms due to their low response frequency which causes them to react relatively slowly [217], [218]. Further, implementing pauses in between power supplied for SMA activation (i.e., alternating between on- and off- modes of SMA actuation, which results in discernable pulsing patterns of compression) has also been shown in this dissertation research to be a particularly effective mode of maintaining comfortable compression and thermal levels for users. As demonstrated in Chapter 6’s SMA-augmented mindful meditation study, the heat generated by the SMAs when electrical power was ‘pulsed’ were almost unanimously percieved as being comfortable and pleasant—hence could also be a promising area of further research and implementation.

As mentioned, SMAs also typically have a one-way shape memory effect. In this dissertation, we introduce the use of braids placed co-axially with the SMA actuators that acts as a passive antagonist element that allows compression-relaxation behaviors with the SMA actuators. However, this solution is currently non-optimized; the braids were selected based on commercial availability. With the current braid stiffness, the relaxation response is slower than that of compression actuation and the braids are prone to large hysteresis. Hence, detailed characterization is required to understand the mechanisms/tradeoffs that affect the performance of these SMA-braid actuator systems, and further investigation on different braid types/parameters are necessary to optimize these SMA-braid actuators [219]. There is room for improvements on the engineering end to understand the mechanical design tradeoffs between the braided sheaths and SMAs, and the pairing of the two results in interesting and beneficial capabilities and could lead to soft robotic innovations that might be relevant to this investigation.

Circling back to a concept that was introduced in Section 2.2.2—the application of compression sensations using SMA actuators are in fact a combination of normal and shear forces. When applied 155 on different body locations, for instance the torso versus shoulders, the amount of shear and normal forces may vary. However, at this current stage, these varying forces cannot be disassociated, and we are unsure if their perception is exclusive from one another. To further understand and characterize the impacts of each of these forces on user perception, psychophysical and experiential studies has to be performed. For instance, with alternative devices only providing normal forces to determine if there are changes in perception when shear forces are added, and on different body locations.

The experience of SMA-driven compression sensations by users in this research work involved indoor, laboratory environments with limited movement by users (all in seated positions), yet the performance and experience of these actuators in the wild when users are moving are unclear (i.e., if the actuators can hold their applied compressive loads and if the varying sensations can be detected by users). Further, these studies also did not tightly control for external factors such as mental stress, attention, context, and they are known to have an effect on mechanosensory thresholds and perception of a stimulus, hence should be considered in future investigations [79].

Applications and Broader Implications

Several user studies relating to the experiential effects of SMA-driven compression and potential applications were investigated in this dissertation work, yet they only scratched the surface of what could be a very promising form of actuation strategy for haptic interaction. The following sections introduce several follow-on, application-driven investigations that should be considered for future work.

7.3.1. Affective Communication

In Chapter 5, we introduced the concept of using warm, compressive haptics for affective communication and studied user expectations of physical and mental strategies used for emotional communication through SMA-based compression sensations. The immediate next step will be to validate and investigate those findings with dyads wearing haptic garments capable of providing those sensations (a study proposal on this topic is presented in Appendix G). While the idea is motivated by the fact that warm compressive haptics is mechanically similar to actual human touch—hence can be viewed as a proxy for direct touch—the effects have not been proven empirically. In other words, further research should be done to investigate the degree to which the 156

SMA-generated system is able to generate touch sensations reliably and believably, or the degree of haptic rendering necessary for the communication to be effective. This could require gathering and converting quantifiable warm, compressive sensations given the communication of various emotions into stimuli that can be replicated to a user wearing a haptic garment and investigate the differences between actual human touch and garment mediated touch. Reliable mapping of the relationships between warm compressive sensations (physically, spatially, and temporally) as an extension of the work presented in Chapter 5 could be extremely useful in various forms of applications including those involving users separated by distance (e.g., communication of emotions between romantic partners or parent-child) or persons who experience visual and/or hearing impairments (e.g., converting facial/emotional expressions into haptic feedback that can be physically experienced and/or felt).

7.3.2. Media Augmentation and Multimodal Interactions

Perhaps one of the most common uses of haptics is in the realm of media augmentation, where haptic technologies are coupled with different media types for the purpose of experience enhancement; this could involve using warm compressive forces to influence the perception of visual media (e.g., photos, videos), auditory media (e.g., music, storytelling), or a combination of both such as movie viewing or gaming experiences; adding the haptic modality to the multisensory experience. This could be further extended to involve immersive AR/VR environments, where we can envision the use of warm, compressive haptics in enhancing presence, immersion, and realism—research in this topic is currently underway [220]. These applications could involve entertainment purposes but also extend to task-driven situations such as collaborative scenarios (i.e., leveraging mediated touch to increase collaboration between partners). Since haptics is used as an ‘augmentation’ tool in these scenarios, they will involve multimodal interactions, coupled with other sensory modalities. Hence, future work should seek to quantify the impacts of the addition of another sensory channel to existing platforms and determine subsequent tradeoffs of the added warm, compressive sensations for varying applications.

7.3.3. Information Transfer and Ambient Notifications

While this dissertation mainly focuses on affective applications, another potential avenue that the SMA actuators could be useful for are ambient awareness/feedback applications. In contrast to vibrotactile feedback that has been found to be ‘annoying’ over lengthy periods of time, 157 compression- or thermal-based haptics have been known to be more pleasant, hence could be effectively used as a form of ambient display, with information presented in the periphery of human attention For instance, to signal ambient factors, temporal progress, or even navigation purposes [18], [23], [24], [35]. While past work has shown promise in this area, the use of SMA actuators in this current configuration (i.e., coiled SMA wires with braided sheaths that allow compressive- relaxation behaviors) has not been done; hence further investigations into different contexts/tasks could shed more insights on the potential of this form of information transfer.

7.3.4. Persuasive Technology

From various clinical interventions (e.g., DTP therapy), there is reason to believe there could be a direct link between compression stimulation and calming effects that could result in attitude/ behavioral changes; our investigations in Chapter 6 through a mindful meditation study lends some support to the above findings. While the positive effects of SMA-driven compression have only been observed within a mindful meditation context (taking into account individual variations in preferences—it is unlikely to be beneficial for individuals who are averse to applied sensations on the body), the next step is to investigate if the benefits can also extend to situations where users experience negative and/or high arousal situations (similar to the contexts at which DTP is applied), in an unobtrusive and implicit manner that requires no conscious actions from users (a study proposal on this topic is presented in Appendix I). In other words, a long-term direction of this form of technology could offer opportunities that extends beyond affect modulation into the realm of persuasive computing or tele-rehabilitation. If coupled with a feedback mechanism that can autonomously detect context or the onset of triggers (e.g., through physiological signals) and apply compressive sensations accordingly, such actuatable compression systems have huge potential for growth as a novel form of persuasive computing.

7.3.5. Dynamic Garments as Expressive Elements

Major portions of this dissertation research investigate compression as experienced sensations that results in changes to perception and/or emotions of a direct user, i.e., they guide experiences inward towards a user. However, dynamically changing garments can also be viewed as a canvas for outwardly displaying status, environmental, or even intimate information such as emotions [221]. In this way, garments that dynamically change, can be used as a form of self-expression or social interaction/communication tool [222]. Currently, the SMA actuators are placed in strategic 158 locations for functional purposes while being as low-profile, discreet as possible. If to be used as a display, these actuators can take on various methods of integration on explicit locations on the body, for example, a tightened-up/compressed position on the shoulders to signify feelings of tension. These actuators can even extend beyond their current configurations that are designed to provide compressive forces to be used as linear actuators positioned parallel to the skin surface, for example, providing the ability to dynamically change the orientation of fabric (e.g., shortening of the sleeves or even provide stroking motions). Such opportunities of blending technology with clothing could have impacts in the future of art and fashion [223], [224].

7.3.6. Ethical Considerations

One final note is that for all the aforementioned research studies produced in this thesis, the garment-mediated compression application was initiated/controlled by a researcher and users have been told what to expect during the study session. However, in some of the applications described, the use cases could involve varying approaches of actuation and there are still plenty of unknowns in regard to how that might change user experiences. For instance, how would the experience change if a wearer knows it was actuated by another person vs. if the device acts autonomously based on sensed information? These experiential, open questions warrant future investigations, but also brings us to an important note on ethical considerations in relation to garment-mediated touch applications. Each of the investigated and suggested (future) applications require varying considerations to potential ethical issues arising from such technologies.

Touch and affective touch (even those mediated through technology) can be intimate and may violate personal or cultural boundaries [10], [15]. When used as a communication system between dyads or emotion regulation/persuasive scenarios, it could also involve privacy and ethical concerns, for instance, what if a receiver does not want to experience the haptic sensations? Or that their emotions are unknowingly/ unwillingly altered during the experience? How should we obtain consent to these varying applications? All of these require serious considerations from designers and researchers, which could involve incorporating safeguards to ensure that users of such systems are protected (e.g., providing autonomy for users to decide if/when/how they experience the haptic sensations).

159

Final Remarks

Taken in aggregate, this research provides insights into how people experience compression on the body given varying stimuli parameters, applications, and contexts. For the first time, we show that SMA actuators are a potential solution to delivering compressive forces on large, distributed areas of the body and explored the design space that influence users’ perception and experience of this form of computer-mediated compression. Through this investigation, we were also able to demonstrate the use of garment-based haptic technologies as tools for interaction and highlight some of the key variables crucial to the experience between humans and wearables. The technology developed and findings gathered from the user studies performed in varying contexts opens up new research opportunities for soft dynamic, wearable systems in affective communication, emotion regulation, immersive environments, and even tele-rehabilitation. Situating this research in the context of a human-machine system, this dissertation contributed to soft robotic technology design, and through the developed systems, provided insights into perceptual and cognitive interactions between technology and human—or user of the system—as well as resulting application spaces the technology could live in; bringing new contributions and understandings to the discipline of human factors. We hope that this work will serve to inform and guide future research opportunities in these areas, and bring some of the ideas on intelligent, novel interfaces discussed in this dissertation closer to reality.

List of Associated Scholarly Work

Refereed Conference Papers and Contributed Journal Publications 1. E. Foo, L. Dunne, B. Holschuh, “User Expectations and Mental Models for Communicating Emotions through Compressive & Warm Affective Garment Actuation”, 2020, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies [Under Review]

2. S. Datar, L. Ferland, E. Foo, M. Kotlyar, B. Holschuh, M. Gini, M. Michalowski and S. Pakhomov, “Measuring Physiological Markers of Stress During Conversational Agent Interactions”, 2021, Studies in Computational Intelligence, Springer International Publishing, Proceedings of the 5th International Workshop on Health Intelligence (W3PHIAI), New York City, NY, USA

3. Miles Priebe, E. Foo, Brad Holschuh, “Shape Memory Alloy Haptic Compression Garment for 160

Media Augmentation in Virtual Reality Environment”, 2020 ACM Symposium on User Interface Software and Technology (UIST), Minneapolis, MN, USA

4. E. Foo, J. Baker, C. Compton, and B. Holschuh, “Soft Robotic Compression Garment to Assist Novice Meditators”, ACM CHI Conference on Human Factors in Computing, CHI’20 Extended Abstracts, Hawaii, USA, © 2020 Association for Computing Machinery. https://doi.org/10.1145/3334480.3382919

5. E. Foo, W.J. Lee, C. Compton, S. Ozbek, and B. Holschuh, “User Experiences of Garment- Based Dynamic Compression for Novel Haptic Applications”, Proceedings of the 23rd International Symposium on Wearable Computers, London, UK, © 2019 Association for Computing Machinery. https://doi.org/ 10.1145/3341163.3347732

6. E. Foo, W.J. Lee, S. Ozbek, C. Compton, and B. Holschuh, “Design and Development of a Garment-based, Dynamic Compression System using Active Materials”, 2019 IEEE World Haptics Conference, Tokyo, Japan.

7. J. W. Lee, E. Foo, S. Ozbek, B. Holschuh, “Investigation of Subjective User Experiences of Applied Passive Compression on Varying Upper Body Locations”, ASME Frontiers in Biomedical Devices, 2019 Design of Medical Devices (DMD) Conference Minneapolis, MN, USA.

8. S. Ozbek, E. Foo, W. J. Lee, N. Schleif, B. Holschuh, “Low-Power, Minimal-Heat Exposure Shape Memory Alloy (SMA) Actuators for On-body Soft Robotics”, ASME Frontiers in Biomedical Devices, 2019 DMD Conference Minneapolis, MN, USA.

9. E. Foo, W.J. Lee, S. Ozbek, and B. Holschuh, “Preliminary Study of Subjective Comfort and Emotional Effects of On-body Compression”, Proceedings of the 22nd International Symposium on Wearable Computers, Singapore, © 2018 Association for Computing Machinery. doi:10.1145/3267242.3267279

Colloquium Presentations at Professional Meetings/Conferences

1. E. Foo, B. Holschuh, “Dynamic Compression in Affective Haptics”, Proceedings of the 2018 ACM International Joint Conference on Pervasive and Ubiquitous Computing, Singapore, © 2018 Association for Computing Machinery. doi: 10.1145/3267305.3267312 161

(UbiComp 2018 Doctoral Colloquium Best Presentation Award)

Design Exhibitions 1. E. Foo, W.J. Lee, S. Ozbek, C. Compton, and B. Holschuh, “Iterative Design and Development of Remotely-Controllable, Dynamic Compression Garment for Novel Haptic Experiences”, Proceedings. of the 23rd International Symposium on Wearable Computers, London, UK, © 2019 Association for Computing Machinery. https://doi.org/10.1145/3341163.3346935

2. E. Foo, W.J. Lee, S. Ozbek, C. Compton, and B. Holschuh, “Garment-based, Computer Mediated Dynamic Compression System using Shape Memory Alloys”, 2019 IEEE World Haptics Conference, Tokyo, Japan. (IEEE World Haptics Conference (WHC) 2019 Best Design Showcase Award)

162

Bibliography

[1] F. McGlone, J. Wessberg, and H. Olausson, “Discriminative and Affective Touch: Sensing and Feeling,” Neuron, vol. 82, no. 4, pp. 737–755, 2014, doi: 10.1016/j.neuron.2014.05.001.

[2] L. K. Case, C. M. Laubacher, H. Olausson, B. Wang, P. A. Spagnolo, and M. C. Bushnell, “Encoding of Touch Intensity But Not Pleasantness in Human Primary Somatosensory Cortex,” J. Neurosci., vol. 36, no. 21, pp. 5850–5860, 2016, doi: 10.1523/JNEUROSCI.1130-15.2016.

[3] G. K. Essick et al., “Quantitative assessment of pleasant touch,” Neurosci. Biobehav. Rev., vol. 34, no. 2, pp. 192–203, 2010, doi: 10.1016/j.neubiorev.2009.02.003.

[4] G. di Pellegrino and E. Làdavas, “Peripersonal space in the brain,” Neuropsychologia, vol. 66, pp. 126–133, 2015, doi: 10.1016/j.neuropsychologia.2014.11.011.

[5] A. Haans and W. IJsselsteijn, “Mediated social touch: A review of current research and future directions,” Virtual Real., vol. 9, no. 2–3, pp. 149–159, 2006, doi: 10.1007/s10055- 005-0014-2.

[6] D. Linden, Touch: The Science of Hand, Heart, and Mind, 1st ed. New York: Penguin Books, 2016.

[7] H. F. Harlow, “The Nature of Love,” 1958.

[8] T. M. Field, “Touch Therapy Effects on Development,” Int. J. Behav. Dev., vol. 22, no. 4, pp. 779–797, 1998.

[9] T. Field, Touch. Cambridge, MA: MIT Press, 2003.

[10] G. Huisman, “Social Touch Technology: A Survey of Haptic Technology for Social Touch,” IEEE Trans. Haptics, vol. 10, no. 3, pp. 391–408, 2017, doi: 10.1109/TOH.2017.2650221.

[11] A. Gallace and C. Spence, “The science of interpersonal touch: An overview,” Neurosci.

163

Biobehav. Rev., vol. 34, no. 2, pp. 246–259, 2010, doi: 10.1016/j.neubiorev.2008.10.004.

[12] A. Crusco and C. Wetzel, “The Midas Touch: The effects of Interpersonal Touch on Restaurant Tipping,” Personal. Soc. Psychol. Rev., vol. 10, no. 4, 1984.

[13] C. Cascio, D. Moore, and F. McGlone, “Social touch and human development.pdf,” Dev. Cogn. Neurosci., vol. 5, no. 11, 2019.

[14] T. Field, “Touch for socioemotional and physical well-being: A review,” Dev. Rev., vol. 30, no. 4, pp. 367–383, 2010, doi: 10.1016/j.dr.2011.01.001.

[15] M. A. Eid and H. Al Osman, “Affective Haptics: Current Research and Future Directions,” Access, IEEE, vol. 4, no. 99, p. 1, 2016, doi: 10.1109/ACCESS.2015.2497316.

[16] K. E. MacLean, O. S. Schneider, and H. Seifi, “Multisensory haptic interactions: understanding the sense and designing for it,” Handb. Multimodal-Multisensor Interfaces Found. User Model. Common Modality Comb. - Vol. 1, pp. 97–142, 2017, doi: 10.1145/3015783.3015788.

[17] A. Delazio et al., “Force Jacket : Pneumatically-Actuated Jacket for Embodied Haptic Experiences,” Conf. Hum. Factors Comput. Syst. - Proc., vol. 2018-April, pp. 1–12, 2018, doi: 10.1145/3173574.3173894.

[18] K. Suhonen, K. Väänänen-Vainio-Mattila, and K. Mäkelä, “User experiences and expectations of vibrotactile, thermal and squeeze feedback in interpersonal communication,” Proc. BCS HCI 2012, pp. 205–214, 2012.

[19] R. Wang and F. Quek, “Touch & talk: Contextualizing remote touch for affective interaction,” TEI’10 - Proc. 4th Int. Conf. Tangible, Embed. Embodied Interact., no. May 2014, pp. 13–20, 2010, doi: 10.1145/1709886.1709891.

[20] M. A. Diego, T. Field, C. Sanders, and M. Hernandez-Reif, “Massage Therapy of Moderate and Light Pressure and Vibrator Effects on Eeg and Heart Rate,” Int. J. Neurosci., vol. 114, no. 1, pp. 31–44, 2004, doi: 10.1080/00207450490249446.

[21] J. R. Hegarty and E. Gale, “Touch as a therapeutic medium for people with challenging

164

behaviours,” Br. J. Learn. Disabil., vol. 24, no. 1, pp. 26–32, 1996, doi: 10.1111/j.1468- 3156.1996.tb00195.x.

[22] T. Grandin, “Calming effects of deep touch pressure in patients with autistic disorder, college students, and animals.,” J. Child Adolesc. Psychopharmacol., vol. 2, no. 1, pp. 63– 72, 1992, doi: 10.1089/cap.1992.2.63.

[23] H. Pohl, F. Hoheisel, and M. Rohs, “Inhibiting Freedom of Movement with Compression Feedback,” Proc. 2017 CHI Conf. Ext. Abstr. Hum. Factors Comput. Syst. - CHI EA ’17, pp. 1962–1969, 2017, doi: 10.1145/3027063.3053081.

[24] G. Chernyshov, F. Cao, B. Tag, G. Liu, C. Caremel, and K. Kunze, “Shape memory alloy wire actuators for soft, wearable haptic devices,” Proc. - Int. Symp. Wearable Comput. ISWC, pp. 112–119, 2018, doi: 10.1145/3267242.3267257.

[25] W. Schiff and M. A. Heller, The Psychology of Touch. New Jersey, 1991.

[26] P. Ramsamy, A. Haffegee, R. Jamieson, and V. Alexandrov, “Using haptics to improve immersion in virtual environments,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 3992 LNCS, pp. 603–609, 2006, doi: 10.1007/11758525_81.

[27] P. M. C. Lemmens, F. Crompvoets, D. Brokken, J. Van Den Eerenbeemd, and G. De Vries, “A body – conforming tactile jacket to enrich movie viewing,” pp. 7–12, 2009.

[28] S. C. Hauser, S. McIntyre, A. Israr, H. Olausson, and G. J. Gerling, “Uncovering Human- to-Human Physical Interactions that Underlie Emotional and Affective Touch Communication,” 2019 IEEE World Haptics Conf. WHC 2019, no. June, pp. 407–412, 2019, doi: 10.1109/WHC.2019.8816169.

[29] S. Mcintyre et al., “Affective touch communication in close adult relationships,” 2019 IEEE World Haptics Conf. WHC 2019, pp. 175–180, 2019, doi: 10.1109/WHC.2019.8816093.

[30] A. Papadopoulou, J. Berry, T. Knight, and R. Picard, Affective sleeve: Wearable materials with haptic action for promoting calmness, vol. 11587 LNCS. Springer International Publishing, 2019.

165

[31] J. Smith and K. MacLean, “Communicating emotion through a haptic link: Design space and methodology,” Int. J. Hum. Comput. Stud., vol. 65, no. 4, pp. 376–387, 2007, doi: 10.1016/j.ijhcs.2006.11.006.

[32] J. N. Bailenson, N. Yee, S. Brave, D. Merget, and D. Koslow, “Virtual interpersonal touch: Expressing and recognizing emotions through haptic devices,” Human-Computer Interact., vol. 22, no. 3, pp. 325–353, 2007, doi: 10.1080/07370020701493509.

[33] S. Yohanan and K. E. MacLean, “Design and assessment of the haptic creature’s affect display,” HRI 2011 - Proc. 6th ACM/IEEE Int. Conf. Human-Robot Interact., pp. 473–480, 2011, doi: 10.1145/1957656.1957820.

[34] H. Pohl, P. Brandes, H. N. Quang, and M. Rohs, “Squeezeback: Pneumatic Compression for Notifications,” Proc. SIGCHI Conf. Hum. Factors Comput. Syst. - CHI ’17, 2017.

[35] A. Gupta, A. A. R. Iruda[1] A. Gupta, A. A. R. Irudayaraj, and R. Balakrishnan, “HapticClench: Investigating Squeeze Sensations using Memory Alloys,” Proc. 30th Annu. ACM Symp. User Interface Softw. Technol. - UIST ’17, pp. 109–117, 2017.yaraj, and R. Balakrishnan, “HapticClench: Investigating Squeeze Sensations using Memory Alloys,” Proc. 30th Annu. ACM Symp. User Interface Softw. Technol. - UIST ’17, pp. 109–117, 2017, doi: 10.1145/3126594.3126598.

[36] S. Singhal, C. Neustaedter, A. N. Antle, and B. Matkin, “Flex-N-Feel: Emotive gloves for physical touch over distance,” CSCW 2017 - Companion 2017 ACM Conf. Comput. Support. Coop. Work Soc. Comput., pp. 37–40, 2017, doi: 10.1145/3022198.3023273.

[37] G. Huisman, A. Darriba Frederiks, B. Van Dijk, D. Hevlen, and B. Krose, “The TaSSt: Tactile sleeve for social touch,” 2013 World Haptics Conf. WHC 2013, pp. 211–216, 2013, doi: 10.1109/WHC.2013.6548410.

[38] M. Zhu et al., “PneuSleeve: In-fabric Multimodal Actuation and Sensing in a Soft, Compact, and Expressive Haptic Sleeve,” To Appear Proc. 2020 CHI Conf. Hum. Factors Comput. Syst. - CHI ’20, 2020.

[39] J. Cha, M. Eid, A. Barghout, A. S. M. M. S. M. M. Rahman, and A. El Saddik, “Hugme: synchronous haptic teleconferencing,” Proc. 17th Int. Conf. Multimed., no. May 2014, pp. 166

1135–1136, 2009, doi: 10.1145/1631272.1631535.

[40] F. Arafsha, K. M. Alam, and A. El Saddik, “EmoJacket: Consumer centric wearable affective jacket to enhance emotional immersion,” 2012 Int. Conf. Innov. Inf. Technol. IIT 2012, pp. 350–355, 2012, doi: 10.1109/INNOVATIONS.2012.6207766.

[41] F. Gemperle, C. Kasabach, J. Stivoric, M. Bauer, and R. Martin, “Design for Wearability,” Ieee Iswc, pp. 116--, 1998.

[42] G. C. Burdea, “Keynote Address: Haptic Feedback for Virtual Reality,” Proc. Int. Work. Virtual prototyping, Laval, Fr., no. May, pp. 87–96, 1999.

[43] O. S. Schneider, H. Seifi, S. Kashani, M. Chun, and K. E. MacLean, “HapTurk: Crowdsourcing affective ratings for vibrotactile icons,” Conf. Hum. Factors Comput. Syst. - Proc., pp. 3248–3260, 2016, doi: 10.1145/2858036.2858279.

[44] H. Culbertson, S. B. Schorr, and A. M. Okamura, “Haptics: The Present and Future of Artificial Touch Sensations,” Annu. Rev. Control Robot. Auton. Syst, vol. 11225, no. 1, pp. 1–12, 2018, doi: 10.1146/annurev-control-060117.

[45] K. E. MacLean, “Haptic Interaction Design for Everyday Interfaces,” Rev. Hum. Factors Ergon., vol. 4, no. 1, pp. 149–194, 2008, doi: 10.1518/155723408x342826.

[46] M. Enriquez, K. MacLean, and C. Chita, “Haptic phonemes: Basic building blocks of haptic communication,” ICMI’06 8th Int. Conf. Multimodal Interfaces, Conf. Proceeding, pp. 302– 309, 2006, doi: 10.1145/1180995.1181053.

[47] C. Pacchierotti, S. Sinclair, M. Solazzi, A. Frisoli, V. Hayward, and D. Prattichizzo, “Wearable haptic systems for the fingertip and the hand: Taxonomy, review, and perspectives,” IEEE Trans. Haptics, vol. 10, no. 4, pp. 580–600, 2017, doi: 10.1109/TOH.2017.2689006.

[48] B. Jackson and L. Rosenberg, “Force Feedback and Medical Simulation,” Interact. Technol. New Paradig. Healthc., pp. 147–151, 1995.

[49] A. C. Castro, J. F. Postigo, and J. Manzano, “Integration of a force feedback joystick with

167

a virtual reality system,” Lat. Am. Appl. Res., vol. 30, no. 2, pp. 171–178, 2000.

[50] D. Tsetserukou, K. Sato, and S. Tachi, “ExoInterfaces: Novel Exosceleton Haptic Interfaces for Virtual Reality, Augmented Sport and Rehabilitation,” Proc. Augment. Human(AH), pp. 1–6, 2010, doi: 10.1145/1785455.1785456.

[51] I. Choi, E. W. Hawkes, D. L. Christensen, C. J. Ploch, and S. Follmer, “Wolverine: A wearable haptic interface for grasping in virtual reality,” IEEE Int. Conf. Intell. Robot. Syst., vol. 2016-Novem, pp. 986–993, 2016, doi: 10.1109/IROS.2016.7759169.

[52] T. H. Massie and J. K. Salisbury, “The PHANTOM Haptic Interface : A Device for Probing Virtual Objects,” ASME Winter Annu. Meet. Symp. Haptic Interfaces Virtual Environ. Teleoperator Syst., vol. 55, pp. 1–6, 1994, doi: http://doi.acm.org/10.1145/1029632.1029682.

[53] L. Brown, “Tactons: Structured Vibrotactile Messages for Non-Visual Information Display,” vol. 28, no. January, p. 228, 2007.

[54] H. Z. Tan and A. Pentland, “Tactual displays for sensory substitution and wearable computers,” ACM SIGGRAPH 2005 Courses on - SIGGRAPH ’05. p. 105, 1997, doi: 10.1145/1198555.1198611.

[55] R. Proctor and T. Van Zandt, Human Factors in Simple and Complex Systems, 2nd ed. CRC Press, 2008.

[56] S. Watkins and L. Dunne, Functional Clothing Design, 1st ed. Fairchild Books, 2015.

[57] A. Okamura, J. Dennerlein, and R. Howe, “Vibration Feedback Models for Virtual Environments,” J. Neurosci., vol. 23, no. 279146–9154, 2003.

[58] G. Frank and S. Carl, “The Cutaneous " Rabbit ": A Perceptual Illusion,” Am. Assoc. Adv. Sci., vol. 178, no. 4057, pp. 178–179, 1972.

[59] A. Israr and I. Poupyrev, “Tactile Brush: Drawing on skin with a tactile grid display,” Conf. Hum. Factors Comput. Syst. - Proc., no. May, pp. 2019–2028, 2011, doi: 10.1145/1978942.1979235.

168

[60] R. Wang, F. Quek, D. Tatar, J. K. S. Teh, and A. D. Cheok, “Keep in touch: Channel, expectation and experience,” Conf. Hum. Factors Comput. Syst. - Proc., pp. 139–148, 2012, doi: 10.1145/2207676.2207697.

[61] A. Israr, S. Zhao, K. Schwalje, R. Klatzky, and J. Lehman, “Feel effects: Enriching storytelling with haptic feedback,” ACM Trans. Appl. Percept., vol. 11, no. 3, pp. 1–17, 2014, doi: 10.1145/2641570.

[62] R. W. Lindeman, R. Page, Y. Yanagida, and J. L. Sibert, “Towards full-body haptic feedback: The design and deployment of a spatialized vibrotactile feedback system,” Proc. ACM Symp. Virtual Real. Softw. Technol. VRST, pp. 146–149, 2004.

[63] F. Arafsha, K. M. Alam, and A. El Saddik, “Design and development of a user centric affective haptic jacket,” Multimed. Tools Appl., vol. 74, no. 9, pp. 3035–3052, 2015, doi: 10.1007/s11042-013-1767-3.

[64] N. Yannier, A. Israr, J. F. Lehman, and R. L. Klatzky, “Feel sleeve: Haptic Feedback to enhance early reading,” Conf. Hum. Factors Comput. Syst. - Proc., vol. 2015-April, no. April, pp. 1015–1024, 2015, doi: 10.1145/2702123.2702396.

[65] J. Ryu and G. J. Kim, “Using a vibro-tactile display for enhanced collision perception and presence,” Proc. ACM Symp. Virtual Real. Softw. Technol. VRST 04, p. 89, 2004, doi: 10.1145/1077534.1077551.

[66] R. W. Lindeman, Y. Yanagida, H. Noma, and K. Hosaka, “Wearable vibrotactile systems for virtual contact and information display,” Virtual Real., vol. 9, no. 2–3, pp. 203–213, 2006, doi: 10.1007/s10055-005-0010-6.

[67] T. Furukawa et al., “Synesthesia wear: Full-body haptic clothing interface based on two- dimensional signal transmission,” SIGGRAPH Asia 2019 Emerg. Technol. SA 2019, pp. 48– 50, 2019, doi: 10.1145/3355049.3360524.

[68] M. J. Hertenstein, D. Keltner, B. App, B. A. Bulleit, and A. R. Jaskolka, “Touch communicates distinct emotions,” Emotion, vol. 6, no. 3, pp. 528–533, 2006, doi: 10.1037/1528-3542.6.3.528.

169

[69] J. Costa, A. T. Adams, M. F. Jung, F. Guimbetière, and T. Choudhury, “EmotionCheck: Leveraging bodily signals and false feedback to regulate our emotions,” UbiComp 2016 - Proc. 2016 ACM Int. Jt. Conf. Pervasive Ubiquitous Comput., no. December 2017, pp. 758– 769, 2016, doi: 10.1145/2971648.2971752.

[70] S. Blairs, S. Slater, and D. J. Hare, “The clinical application of deep touch pressure with a man with autism presenting with severe anxiety and challenging behaviour,” Br. J. Learn. Disabil., vol. 35, no. 4, pp. 214–220, 2007, doi: 10.1111/j.1468-3156.2007.00452.x.

[71] M. J. Enriquez and K. E. MacLean, “The hapticon editor: A tool in support of haptic communication research,” Proc. - 11th Symp. Haptic Interfaces Virtual Environ. Teleoperator Syst. HAPTICS 2003, pp. 356–362, 2003, doi: 10.1109/HAPTIC.2003.1191310.

[72] F. “Floyd” F. Mueller, F. Vetere, M. R. Gibbs, J. Kjeldskov, S. Pedell, and S. Howard, “Hug over a distance,” CHI ’05 Ext. Abstr. Hum. factors Comput. Syst. - CHI ’05, p. 1673, 2005, doi: 10.1145/1056808.1056994.

[73] A. Chang, S. O’Modhrain, R. Jacob, E. Gunther, and H. Ishii, “ComTouch,” Proc. Conf. Des. Interact. Syst. Process. Pract. methods, Tech. - DIS ’02, p. 312, 2002, doi: 10.1145/778712.778755.

[74] J. K. S. Teh, Z. Tsai, J. T. K. V. Koh, and A. D. Cheok, “Mobile implementation and user evaluation of the Huggy Pajama system,” Haptics Symp. 2012, HAPTICS 2012 - Proc., no. March 2012, pp. 471–478, 2012, doi: 10.1109/HAPTIC.2012.6183833.

[75] S. Yarosh, B. Unver, X. Wang, Y. Yao, A. Campbell, and B. Holschuh, “SqueezeBands : Mediated Social Touch Using Shape Memory Alloy Actuation,” vol. 1, no. November, pp. 1–18, 2017.

[76] N. Gleveckas-Martens, “Somatosensory System Anatomy,” MedScape, 2013. [Online]. Available: http://emedicine.medscape.com/article/1948621-overview.

[77] S. Bhatnagar, P. Sabatini, A. Klingler, A. Noplock, and S. Glover, “Somatosensory System,” Neurosci. Study Commun. Disord., pp. 164–84, 2008.

170

[78] S. Stranding, Gray’s Anatomy, 40th Editi. Elseiver, 2008.

[79] D. Purves, G. Augustine, and D. Fitzpatrick, Neuroscience, 2nd ed. Sunderland (MA), 2001.

[80] N. Cauna and L. L. Ross, “The fine structure of Meissner’s touch corpuscles of human fingers.,” J. Biophys. Biochem. Cytol., vol. 8, no. 2, pp. 467–82, 1960, doi: 10.1083/jcb.8.2.467.

[81] M. Paré, R. Elde, J. E. Mazurkiewicz, a M. Smith, and F. L. Rice, “The Meissner corpuscle revised: a multiafferented mechanoreceptor with nociceptor immunochemical properties.,” J. Neurosci., vol. 21, no. 18, pp. 7236–7246, 2001.

[82] M. J. Welsh, M. P. Price, and J. Xie, “Biochemical basis of touch perception: Mechanosensory function of degenerin/epithelial Na + channels,” J. Biol. Chem., vol. 277, no. 4, pp. 2369–2372, 2002, doi: 10.1074/jbc.R100060200.

[83] K. O. Johnson, T. Yoshioka, and F. Vega Bermudez, “Tactile functions of mechanoreceptive afferents innervating the hand,” J. Clin. Neurophysiol., vol. 17, no. 6, pp. 539–558, 2000, doi: 10.1097/00004691-200011000-00002.

[84] R. S. Johansson and J. R. Flanagan, “Coding and use of tactile signals from the fingertips in object manipulation tasks,” Nat. Rev. Neurosci., vol. 10, no. 5, pp. 345–359, 2009, doi: 10.1038/nrn2621.

[85] G. Gescheider, Psychophysics: the fundamentals, 3rd ed. Lawrence Erlbaum Associates., 1997.

[86] R. S. Johansson and A. B. Vallbo, “Detection of tactile stimuli. Thresholds of afferent units related to psychophysical thresholds in the human hand.,” J. Physiol., vol. 297, pp. 405–22, 1979, doi: 0022-3751/79/6320-0181.

[87] J. Tong, O. Mao, and D. Goldreich, “Two-point orientation discrimination versus the traditional two-point test for tactile spatial acuity assessment,” Front. Hum. Neurosci., vol. 7, no. September, pp. 1–11, 2013, doi: 10.3389/fnhum.2013.00579.

[88] J. A. Bell-Krotoski, E. E. Fess, J. H. Figarola, and D. Hiltz, “Threshold Detection and

171

Semmes-Weinstein Monofilaments,” J. Hand Ther., vol. 8, no. 2, pp. 155–162, 1995, doi: 10.1016/S0894-1130(12)80314-0.

[89] E. Weber, E.H. Weber on the Tactile Senses. Erlbaum (UK): Taylor & Francis, 1996.

[90] F. Mancini et al., “Whole-body mapping of spatial acuity for pain and touch,” Ann. Neurol., vol. 75, no. 6, pp. 917–924, 2014, doi: 10.1002/ana.24179.

[91] T. Mitsuda, “Pseudoforce Display that Applies Pressure to the Forearms,” Presence Teleoperators Virtual Environ., vol. 22, no. 3, pp. 191–201, 2013, doi: 10.1162/PRES.

[92] I. Morrison, L. S. Löken, and H. Olausson, “The skin as a social organ,” Exp. Brain Res., vol. 204, no. 3, pp. 305–314, 2010, doi: 10.1007/s00221-009-2007-y.

[93] K. Dharani, “The Perceptual Neurons,” in The biology of thought : a neuronal mechanism in the generation of thought - a new molecular model, London, England ; San Diego, California : AP, 2015.

[94] D. M. Ellingsen, S. Leknes, G. Løseth, J. Wessberg, and H. Olausson, “The neurobiology shaping affective touch: Expectation, motivation, and meaning in the multisensory context,” Front. Psychol., vol. 6, no. JAN, pp. 1–16, 2016, doi: 10.3389/fpsyg.2015.01986.

[95] L. S. Löken, M. Evert, and J. Wessberg, “Pleasantness of touch in human glabrous and hairy skin: Order effects on affective ratings,” Brain Res., vol. 1417, pp. 9–15, 2011, doi: 10.1016/j.brainres.2011.08.011.

[96] M. Bianchi, “A fabric-based approach for wearable haptics,” Electron., vol. 5, no. 3, pp. 1– 14, 2016, doi: 10.3390/electronics5030044.

[97] H. Pohl, F. Hoheisel, and M. Rohs, “Inhibiting freedom of movement with compression feedback,” Conf. Hum. Factors Comput. Syst. - Proc., vol. Part F1276, pp. 1962–1969, 2017, doi: 10.1145/3027063.3053081.

[98] M. A. Baumann, K. E. MacLean, T. W. Hazelton, and A. McKay, “Emulating human attention-getting practices with wearable haptics,” 2010 IEEE Haptics Symp. HAPTICS 2010, pp. 149–156, 2010, doi: 10.1109/HAPTIC.2010.5444662.

172

[99] S. Jones and E. Yarbrough, “A Naturalistic Study of the Meanings of Touch,” Commun. Monogr., vol. 52, 1985.

[100] G. Huisman, A. Darriba Frederiks, B. Van Dijk, D. Hevlen, and B. Krose, “The TaSSt: Tactile sleeve for social touch,” 2013 World Haptics Conf. WHC 2013, pp. 211–216, 2013, doi: 10.1109/WHC.2013.6548410.

[101] D. Tsetserukou, “HaptiHug: A novel haptic display for communication of hug over a distance,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 6191 LNCS, no. PART 1, pp. 340–347, 2010, doi: 10.1007/978-3- 642-14064-8_49.

[102] J. K. S. Teh, Y. Choi, A. D. Cheok, V. Thuong, R. L. Peiris, and S. Lai, “Huggy Pajama: a mobile parent and child hugging communication system,” Proc. 7th Int. Conf. Interact. Des. Child., no. January, pp. 250–257, 2008, doi: 10.1145/1551788.1551861.

[103] D. Tsetserukou and A. Neviarouskaya, “Emotion telepresence: Emotion augmentation through affective haptics and visual stimuli,” J. Phys. Conf. Ser., vol. 352, no. 1, 2012, doi: 10.1088/1742-6596/352/1/012045.

[104] A. Mazzoni and N. Bryan-Kinns, “Mood Glove: A haptic wearable prototype system to enhance mood music in film,” Entertain. Comput., vol. 17, pp. 9–17, 2016, doi: 10.1016/j.entcom.2016.06.002.

[105] S. Zhao, O. Schneider, R. Klatzky, J. Lehman, and A. Israr, “FeelCraft: Crafting tactile experiences for media using a feel effect library,” UIST 2014 - Adjun. Publ. 27th Annu. ACM Symp. User Interface Softw. Technol., no. 1, pp. 51–52, 2014, doi: 10.1145/2658779.2659109.

[106] B. Fogg, “Persuasive computers: Perspectives and research directions,” Proc. SIGCHI Conf. Hum. Factors Comput. Syst., no. April, pp. 225–232, 1998, doi: 10.1145/274644.274677.

[107] Y. Gaffary and A. Lécuyer, “The Use of Haptic and Tactile information in the Car to improve Driving Safety : A Review of Current Technologies,” vol. 5, no. March, pp. 1–11, 2018, doi: 10.3389/fict.2018.00005.

173

[108] J. Burgoon, “Interpersonal Expectations, Expectancy Violations, and Emotional Communication,” Journal of Language and Social Psychology, vol. 12, no. 1. pp. 30–48, 1993, doi: 10.1089/cpb.2006.9993.

[109] J. D. Fisher, M. Rytting, and R. Heslin, “Hands Touching Hands : Affective and Evaluative Effects of an Interpersonal Touch Author ( s ): Jeffrey D . Fisher , Marvin Rytting and Richard Heslin Published by : American Sociological Association Stable URL : http://www.jstor.org/stable/3033506 Access,” Sociometry, vol. 39, no. 4, pp. 416–421, 1976.

[110] J. Hornik and S. Ellis, “Strategies to secure compliance for a mall intercept interview,” Public Opin. Q., vol. 52, no. 4, pp. 77–79, 1988.

[111] N. Gueguen and J. Fischer-Lokou, “Another Evaluation Of Touch and Helping Behavior’,” Psychol. Reports O Psychol. Reports, vol. 92, pp. 62–64, 2003.

[112] R. Adams, B. White, and C. Beckett, “The effects of massage therapy on pain management in the acute care setting,” Int. J. Ther. Massage Bodyw. Res. Educ. Pract., vol. 3, no. 1, pp. 4–11, 2010, doi: 10.3822/ijtmb.v3i1.54.

[113] E. D. E. Papathanassoglou and M. D. A. Mpouzika, “Interpersonal Touch: Physiological Effects in Critical Care,” Biol. Res. Nurs., vol. 14, no. 4, pp. 431–443, 2012, doi: 10.1177/1099800412451312.

[114] E. Bush, “The Use of Human Touch to Improve the Well-Being of Older Adults: A Holistic Nursing Intervention,” J. Holist. Nurs., vol. 19, no. 3, pp. 256–270, 2001, doi: 10.1177/089801010101900306.

[115] N. L. Vandenberg, “The Use of a Weighted Vest to Increase On-Task Behavior in Children with Attention Difficulties,” Am. J. Occup. Ther., vol. 55, no. 6, pp. 621–628, 1998.

[116] T. N. Davis et al., “The effects of a weighted vest on aggressive and self-injurious behavior in a child with autism,” Dev Neurorehabil, vol. 16, no. 3, pp. 210–215, 2013, doi: 10.3109/17518423.2012.753955.

[117] S. S. Roley, Z. Mailloux, H. Miller-Kuhaneck, and T. Glennon, “Understanding Ayres’ Sensory Integration,” AOTA Contin. Educ., vol. 12, no. 17, pp. 1–8, 2007.

174

[118] A. J. Ayres, Sensory integration and learning disorders. 1972.

[119] A. A. Stanley and K. J. Kuchenbecker, “Design of body-grounded tactile actuators for playback of human physical contact,” 2011 IEEE World Haptics Conf., pp. 563–568, 2011, doi: 10.1109/WHC.2011.5945547.

[120] E. Morrison, E, “a Review of Research on the Use of Weighted Vests With Children on the Autism Spectrum,” Education, vol. 127, no. 3, pp. 323–328, 2004.

[121] M. J. Brennan and L. T. Miller, “Overview of treatment options and review of the current role and use of compression garments, intermittent pumps, and exercise in the management of lymphedema.,” Cancer, vol. 83, no. 12 Suppl American, pp. 2821–2827, 1998, doi: 10.1002/(SICI)1097-0142(19981215)83:12B+<2821::AID-CNCR33>3.0.CO;2-G [pii].

[122] My T Jacket, “T Jacket.” [Online]. Available: https://www.mytjacket.com/the-science- behind-it.html.

[123] B. Holschuh and D. Newman, “Two-spring model for active compression textiles with integrated NiTi coil actuators,” Smart Mater. Struct., vol. 24, no. 3, p. 35011, 2015, doi: 10.1088/0964-1726/24/3/035011.

[124] E. Pezent et al., “Tasbi: Multisensory Squeeze and Vibrotactile Wrist Haptics for Augmented and Virtual Reality,” 2019 IEEE World Haptics Conf. WHC 2019, pp. 1–6, 2019, doi: 10.1109/WHC.2019.8816098.

[125] F. Chinello, M. Aurilio, C. Pacchierotti, D. Prattichizzo, and D. Prattichizzo, “The HapBand: A cutaneous device for remote tactile interaction,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 8618, pp. 284–291, 2014, doi: 10.1007/978-3-662-44193-0_36.

[126] D. Tsetserukou and A. Neviarouskaya, “iFeel_IM!: Augmenting Emotions during Online Communication,” IEEE Comput. Graph. Appl., vol. 30, no. 5, pp. 72–80, 2010, doi: 10.1109/MCG.2010.88.

[127] E. M. Young, A. H. Memar, P. Agarwal, and N. Colonnese, “Bellowband: A Pneumatic Wristband for Delivering Local Pressure and Vibration,” 2019 IEEE World Haptics Conf.

175

WHC 2019, pp. 55–60, 2019, doi: 10.1109/WHC.2019.8816075.

[128] M. Raitor, J. M. Walker, A. M. Okamura, and H. Culbertson, “WRAP: Wearable, restricted- aperture pneumatics for haptic guidance,” Proc. - IEEE Int. Conf. Robot. Autom., pp. 427– 432, 2017, doi: 10.1109/ICRA.2017.7989055.

[129] S. Reynolds, S. J. Lane, and B. Mullen, “Effects of Deep Pressure Stimulation on Physiological Arousal,” Am. J. Occup. Ther., vol. 69, 2015.

[130] L. He, C. Xu, D. Xu, and R. Brill, “PneuHaptic: Delivering Haptic Cues with a Pneumatic Armband,” Proc. 2015 ACM Int. Symp. Wearable Comput. - ISWC ’15, pp. 47–48, 2015, doi: 10.1145/2802083.2802091.

[131] K. Ueda, T. Terada, and M. Tsukamoto, “Haptic Feedback Method using Deformation of Clothing,” pp. 84–93, 2019, doi: 10.1145/3365921.3365933.

[132] N. Alhuda Hamdan, A. Wagner, S. Voelker, J. Steimle, and J. Borchers, “Springlets: Expressive, flexible and silent on-skin tactile interfaces,” Conf. Hum. Factors Comput. Syst. - Proc., pp. 1–14, 2019, doi: 10.1145/3290605.3300718.

[133] J. C. Duvall, L. E. Dunne, N. Schleif, and B. Holschuh, “Active ‘Hugging’ Vest for Deep Touch Pressure Therapy,” Proc. 2016 ACM Int. Jt. Conf. Pervasive Ubiquitous Comput. Adjun., pp. 458–463, 2016, doi: 10.1145/2968219.2971344.

[134] M. J. Hertenstein, R. Holmes, M. McCullough, and D. Keltner, “The Communication of Emotion via Touch,” Emotion, vol. 9, no. 4, pp. 566–573, 2009, doi: 10.1037/a0016108.

[135] C. Abras, D. Maloney-Krichmar, and J. Preece, User-Centered Design. 2004.

[136] V. Ferraro and S. Ugur, “Designing wearable technologies through a user centered approach,” DPPI’11 - Des. Pleasurable Prod. Interfaces, Proc., no. c, 2011, doi: 10.1145/2347504.2347510.

[137] J. W. Lee, E. Foo, S. Ozbek, and B. Holschuh, “Investigation of Subjective User Experiences of Applied Passive Compression on Varying Upper Body Locations,” Proc. Des. Med. Devices Conf., pp. 16–18, 2019.

176

[138] P. C. Wright and A. F. Monk, “The use of think-aloud evaluation methods in design,” ACM SIGCHI Bull., vol. 23, no. 1, pp. 55–57, 1991, doi: 10.1145/122672.122685.

[139] H. Meng and G. Li, “A review of stimuli-responsive shape memory polymer composites,” Polymer (Guildf)., vol. 54, no. 9, pp. 2199–2221, 2013, doi: 10.1016/j.polymer.2013.02.023.

[140] W. M. Huang, Z. Ding, C. C. Wang, J. Wei, Y. Zhao, and H. Purnawali, “Shape memory materials,” Mater. Today, vol. 13, no. 7–8, pp. 54–61, 2010, doi: 10.1016/S1369- 7021(10)70128-0.

[141] A. O’Halloran, F. O’Malley, and P. McHugh, “A review on dielectric elastomer actuators, technology, applications, and challenges,” J. Appl. Phys., vol. 104, no. 7, 2008, doi: 10.1063/1.2981642.

[142] T. Carter, S. A. Seah, B. Long, B. Drinkwater, and S. Subramanian, “UltraHaptics : Multi- Point Mid-Air Haptic Feedback for Touch Surfaces,” Proc. UIST 2013, pp. 505–514, 2013, doi: 10.1145/2501988.2502018.

[143] G. Andrikopoulos, G. Nikolakopoulos, and S. Manesis, “2011 A survey on applications of pneumatic artificial muscles,” no. July 2015, 2011.

[144] F. Gemperle, N. Ota, and D. Siewiorek, “Design of a wearable tactile display,” Proc. Fifth Int. Symp. Wearable Comput., pp. 5–12, 2001, doi: 10.1109/ISWC.2001.962082.

[145] F. Arafsha, “Smart Clothes as a Tangible User Interface to Affect Human Emotions using Haptic Actuators,” no. January 2012, 2012, doi: 10.20381/ruor-5388.

[146] A. Al Maimani and A. Roudaut, “Frozen suit: Toward a changeable stiffness suit and its application for haptic games,” Conf. Hum. Factors Comput. Syst. - Proc., vol. 2017-May, pp. 2440–2448, 2017, doi: 10.1145/3025453.3025655.

[147] A. A. Stanley, J. C. Gwilliam, and A. M. Okamura, “Haptic jamming: A deformable geometry, variable stiffness tactile display using pneumatics and particle jamming,” 2013 World Haptics Conf. WHC 2013, pp. 25–30, 2013, doi: 10.1109/WHC.2013.6548379.

[148] B. Holschuh, E. Obropta, L. Buechley, and D. Newman, “Materials and Textile Architecture

177

Analyses for Mechanical Counter-Pressure Space Suits using Active Materials,” AIAA Sp. 2012 Conf. Expo., no. September, pp. 1–17, 2012, doi: 10.2514/6.2012-5206.

[149] H. Fruhstorfer, U. Lindblom, and W. C. Schmidt, “Method for quantitative estimation of thermal thresholds in patients,” J Neurol Neurosurg Psychiatry, vol. 39, no. 11, pp. 1071– 1075, 1976, doi: 10.1136/jnnp.39.11.1071.

[150] D. C. Lagoudas, Shape Memory Alloys, vol. 1. 2008.

[151] A. Rao, A. R. Srinivasa, and J. N. Reddy, Design of Shape Memory Alloy (SMA) Actuators. 2015.

[152] S. Doraiswamy, A. Rao, and A. R. Srinivasa, “Combining thermodynamic principles with Preisach models for superelastic shape memory alloy wires,” Smart Mater. Struct., vol. 20, no. 8, 2011, doi: 10.1088/0964-1726/20/8/085032.

[153] D. Carnelli, G. Pennati, T. Villa, L. Baglioni, B. Reimers, and F. Migliavacca, “Mechanical Properties of Open-Cell, Self-Expandable Shape Memory Alloy Carotid Stents,” Artif. Organs, vol. 35, no. 1, pp. 74–80, 2011, doi: 10.1111/j.1525-1594.2010.01018.x.

[154] A. T. Tung, B. H. Park, G. Niemeyer, and D. H. Liang, “Laser-machined shape memory alloy actuators for active catheters,” IEEE/ASME Trans. Mechatronics, vol. 12, no. 4, pp. 439–446, 2007, doi: 10.1109/TMECH.2007.901926.

[155] S. Wu, X. Liu, K. W. K. Yeung, Z. S. Xu, C. Y. Chung, and P. K. Chu, “Wear Properties of Porous NiTi Orthopedic Shape Memory Alloy,” J. Mater. Eng. Perform., vol. 21, no. 12, pp. 2622–2627, 2012, doi: 10.1007/s11665-012-0392-z.

[156] M. Kheirikhah, S. Rabiee, and M. Edalat, “A Review of Shape Memory Alloy Actuators in Robotics,” Rob. 2010 Robot Soccer World Cup XIV, pp. 206–217, 2011.

[157] A. Villoslada, A. Flores-Caballero, D. Copaci, D. Blanco, and L. Moreno, “High- displacement fast-cooling flexible Shape Memory Alloy actuator: Application to an anthropomorphic robotic hand,” IEEE-RAS Int. Conf. Humanoid Robot., vol. 2015-Febru, pp. 27–32, 2015, doi: 10.1109/HUMANOIDS.2014.7041333.

178

[158] J. Duvall et al., “The design and development of active compression garments for orthostatic intolerance,” Front. Biomed. Devices, BIOMED - 2017 Des. Med. Devices Conf. DMD 2017, pp. 2017–2018, 2017, doi: 10.1115/DMD2017-3480.

[159] P. K. Kumar and D. C. Lagoudas, Shape Memory Alloys, vol. 1. 2008.

[160] B. Holschuh, E. Obropta, and D. Newman, “Low spring index NiTi coil actuators for use in active compression garments,” IEEE/ASME Trans. Mechatronics, vol. 20, no. 3, pp. 1264– 1277, 2015, doi: 10.1109/TMECH.2014.2328519.

[161] B. Holschuh, E. Obropta, and D. Newman, “Low Spring Index NiTi Coil Actuators,” pp. 1–14, 2014.

[162] R. Granberry, N. Ciavarella, R. Pettys-Baker, M. E. Berglund, and B. Holschuh, “No- power-required, touch-activated compression garments for the treatment of pots,” Front. Biomed. Devices, BIOMED - 2018 Des. Med. Devices Conf. DMD 2018, no. 4, pp. 1–5, 2018, doi: 10.1115/DMD2018-6886.

[163] M. Bianchi et al., “Design and preliminary affective characterization of a novel fabric-based tactile display,” IEEE Haptics Symp. HAPTICS, pp. 591–596, 2014, doi: 10.1109/HAPTICS.2014.6775522.

[164] E. Foo, J. W. Lee, S. Ozbek, and B. Holschuh, “Preliminary Study of the Subjective Comfort and Emotional Effects of On-Body Compression,” 2018 Int. Symp. Wearable Comput., 2018.

[165] J. C. Duvall, “Bear Hug: The Design and Development of an Active Deep Touch Pressure Garment for Sensory Processing Disorder,” University of Minnesota, 2017.

[166] C. J. Smith and G. Havenith, “Body mapping of sweating patterns in male athletes in mild exercise-induced hyperthermia,” Eur. J. Appl. Physiol., vol. 111, no. 7, pp. 1391–1404, 2011, doi: 10.1007/s00421-010-1744-8.

[167] E. Foo, J. W. Lee, S. Ozbek, C. Compton, and B. Holschuh, “Iterative Design and Development of Remotely-Controllable , Dynamic Compression Garment for Novel Haptic Experiences,” Int. Symp. Wearable Comput. Des. Exhib., 2019.

179

[168] E. Foo, J. W. Lee, C. Compton, S. Ozbek, and B. T. Holschuh, “User Experiences of Garment - Based Dynamic Compression for Novel Haptic Applications,” 2019 Int. Symp. Wearable Comput., 2019.

[169] L. MacIntyre, “New calibration method for I-scan sensors to enable the precise measurement of pressures delivered by ‘pressure garments,’” Burns, vol. 37, no. 7, pp. 1174–1181, 2011, doi: 10.1016/j.burns.2011.06.008.

[170] L. Macintyre, “Designing pressure garments capable of exerting specific pressures on limbs,” Burns, vol. 33, no. 5, pp. 579–586, 2007, doi: 10.1016/j.burns.2006.10.004.

[171] B. N. Lang, P.J., Bradley, M.M., & Cuthbert, “International affective picture system (IAPS): Affective ratings of pictures and instruction manual.,” 2008.

[172] A. V Cardello, C. Winterhalter, and H. G. Schutz, “Predicting the Handle and Comfort of Military Clothing Fabrics from Sensory and Instrumental Data,” Text. Res. J., vol. 73, no. 3, pp. 221–237, 2003.

[173] M. Bradley and P. J. Lang, “Measuring Emotion: The Self Assessment Manikin and the Semantic Differential,” vol. 25, no. I, 1994.

[174] S. Ozbek, E. Foo, J. W. Lee, N. Schleif, and B. Holschuh, “Low-Power, Minimal Heat Exposure Shape Memory Alloy (SMA) Actuators for On-body Soft Robotics,” Proc. 2019 Des. Med. Devices Conf., pp. 16–18, 2019.

[175] G. Huisman, A. D. Frederiks, and A. Darriba Frederiks, “Towards tactile expressions of emotion through mediated touch,” CHI ’13 Ext. Abstr. Hum. Factors Comput. Syst. - CHI EA ’13, vol. 2013-April, p. 1575, 2013, doi: 10.1145/2468356.2468638.

[176] D. Tsetserukou and A. Neviarouskaya, “World’s first wearable humanoid robot that augments our emotions,” ACM Int. Conf. Proceeding Ser., no. May, 2010, doi: 10.1145/1785455.1785463.

[177] G. Wilson, D. Dobrev, and S. A. Brewster, “Hot under the collar: Mapping thermal feedback to dimensional models of emotion,” Conf. Hum. Factors Comput. Syst. - Proc., no. June, pp. 4838–4849, 2016, doi: 10.1145/2858036.2858205.

180

[178] G. Wilson and S. A. Brewster, “Multi-Moji: Combining thermal, vibrotactile & visual stimuli to expand the affective range of feedback,” Conf. Hum. Factors Comput. Syst. - Proc., vol. 2017-May, no. January, pp. 1743–1755, 2017, doi: 10.1145/3025453.3025614.

[179] P. Ekman, “Basic Emotions,” Handbook of Cognition and Emotion. pp. 1–6, 1999, doi: 10.1007/978-3-319-28099-8_495-1.

[180] P. Ekman, “Are there Basic Emotions,” Psychol. Rev., vol. 99, no. 3, p. 550, 1992, doi: 10.4081/jear.2011.169.

[181] L. Nummenmaa, E. Glerean, R. Hari, and J. K. Hietanen, “Bodily maps of emotions,” Proc. Natl. Acad. Sci. U. S. A., vol. 111, no. 2, pp. 646–651, 2014, doi: 10.1073/pnas.1321664111.

[182] S. Volynets, E. Glerean, J. K. Hietanen, R. Hari, and L. Nummenmaa, “Bodily Maps of Emotions Are Culturally Universal,” Emotion, pp. 1–10, 2019, doi: 10.1037/emo0000624.

[183] J. J. Gross, “Emotion regulation: Past, present, future,” Cogn. Emot., vol. 13, no. 5, pp. 551– 573, 1999, doi: 10.1080/026999399379186.

[184] J. J. Gross, “Emotion regulation: Affective, cognitive, and social consequences,” Psychophysiology, vol. 39, no. 3, pp. 281–291, 2002, doi: 10.1017/S0048577201393198.

[185] C. B. Menezes, M. G. Pereira, and L. Bizarro, “Sitting and silent meditation as a strategy to study emotion regulation,” Psychol. Neurosci., vol. 5, no. 1, pp. 27–36, 2012, doi: 10.3922/j.psns.2012.1.05.

[186] T. Esch, “The neurobiology of meditation,” Medit. – Neurosci. Approaches Philos. Implic. Stud. Neurosci. Conscious. Spiritual., vol. 2, pp. 153–173, 2014, doi: 10.1007/978-3-319- 01634-4.

[187] P. R. Goldin and J. J. Gross, “Effects of Mindfulness-Based Stress Reduction (MBSR) on Emotion Regulation in Social Anxiety Disorder,” Emotion, vol. 10, no. 1, pp. 83–91, 2010, doi: 10.1037/a0018441.

[188] C. D. Roquet and C. Sas, “Evaluating mindfulness meditation apps,” Conf. Hum. Factors Comput. Syst. - Proc., vol. 2018-April, pp. 1–6, 2018, doi: 10.1145/3170427.3188616.

181

[189] J. Vidyarthi and B. E. Riecke, “Interactively mediating experiences of mindfulness meditation,” Int. J. Hum. Comput. Stud., vol. 72, no. 8–9, pp. 674–688, 2014, doi: 10.1016/j.ijhcs.2014.01.006.

[190] I. Plaza, M. M. P. Demarzo, P. Herrera-Mercadal, and J. García-Campayo, “Mindfulness- based mobile applications: Literature review and analysis of current features,” J. Med. Internet Res., vol. 15, no. 11, 2013, doi: 10.2196/mhealth.2733.

[191] A. Ghandeharioun and R. Picard, “BrightBeat: Effortlessly influencing breathing for cultivating calmness and focus,” Conf. Hum. Factors Comput. Syst. - Proc., vol. Part F1276, pp. 1624–1631, 2017, doi: 10.1145/3027063.3053164.

[192] C. Kelling, D. Pitaro, and J. Rantala, “Good vibes: The impact of haptic patterns on stress levels,” Acad. 2016 - Proc. 20th Int. Acad. Mindtrek Conf., no. October, pp. 130–136, 2016, doi: 10.1145/2994310.2994368.

[193] C. K. Peng et al., “Heart rate dynamics during three forms of meditation,” Int. J. Cardiol., vol. 95, no. 1, pp. 19–27, 2004, doi: 10.1016/j.ijcard.2003.02.006.

[194] W. J. Elliott et al., “Graded Blood Pressure Reduction in Hypertensive Outpatients Associated With Use of a Device to Assist With Slow Breathing,” J. Clin. Hypertens., vol. 6, no. 10, pp. 553–559, 2004, doi: 10.1111/j.1524-6175.2004.03553.x.

[195] J. D. Nash and A. Newberg, “Toward a unifying taxonomy and definition for meditation,” Front. Psychol., vol. 4, no. NOV, pp. 1–18, 2013, doi: 10.3389/fpsyg.2013.00806.

[196] A. Lutz, A. P. Jha, J. D. Dunne, and C. D. Saron, “Investigating the Phenomenologial Matrix of Mindfulness-related Practies from a Neurocognitive Perspective,” Author Manuscr., vol. 70, no. 7, pp. 632–658, 2015, doi: 10.1037/a0039585.

[197] D. B. Levinson, J. Smallwood, and R. J. Davidson, “The Persistence of Thought: Evidence for a Role of Working Memory in the Maintenance of Task-Unrelated Thinking,” Psychol. Sci., vol. 23, no. 4, pp. 375–380, 2012, doi: 10.1177/0956797611431465.

[198] W. Hasenkamp, C. D. Wilson-Mendenhall, E. Duncan, and L. W. Barsalou, “Mind wandering and attention during focused meditation: A fine-grained temporal analysis of

182

fluctuating cognitive states,” Neuroimage, vol. 59, no. 1, pp. 750–760, 2012, doi: 10.1016/j.neuroimage.2011.07.008.

[199] G. K. Pal, S. Velkumary, and A. Madanmohan, “Effect of short-term practice of breathing exercises on autonomic functions in normal human volunteers,” Indian J. Med. Res., vol. 120, no. 2, pp. 115–121, 2004.

[200] M. Feuille and K. Pargament, “Pain, mindfulness, and spirituality: A randomized controlled trial comparing effects of mindfulness and relaxation on pain-related outcomes in migraineurs,” J. Health Psychol., vol. 20, no. 8, pp. 1090–1106, 2015, doi: 10.1177/1359105313508459.

[201] H. Cho, S. Ryu, J. Noh, and J. Lee, “The effectiveness of daily mindful breathing practices on test anxiety of students,” PLoS One, vol. 11, no. 10, pp. 1–10, 2016, doi: 10.1371/journal.pone.0164822.

[202] J. A. Brewer, P. D. Worhunsky, J. R. Gray, Y. Y. Tang, J. Weber, and H. Kober, “Meditation experience is associated with differences in default mode network activity and connectivity,” Proc. Natl. Acad. Sci. U. S. A., vol. 108, no. 50, pp. 20254–20259, 2011, doi: 10.1073/pnas.1112029108.

[203] D. B. Wolf and N. Abell, “Examining the effects of meditation techniques on psychosocial functioning,” Res. Soc. Work Pract., vol. 13, no. 1, pp. 27–42, 2003, doi: 10.1177/104973102237471.

[204] A. T. Adams, J. Costa, M. F. Jung, and T. Choudhury, “Mindless computing: Designing technologies to subtly influence behavior,” UbiComp 2015 - Proc. 2015 ACM Int. Jt. Conf. Pervasive Ubiquitous Comput., pp. 719–730, 2015, doi: 10.1145/2750858.2805843.

[205] J. Holt-Lunstad, W. A. Birmingham, and K. C. Light, “Influence of a ‘warm touch’ support enhancement intervention among married couples on ambulatory blood pressure, oxytocin, alpha amylase, and cortisol,” Psychosom. Med., vol. 70, no. 9, pp. 976–985, 2008, doi: 10.1097/PSY.0b013e318187aef7.

[206] R. I. Ramos-Garcia, F. Da Silva, Y. Kondi, E. Sazonov, and L. E. Dunne, “Analysis of a coverstitched stretch sensor for monitoring of breathing,” Proc. Int. Conf. Sens. Technol. 183

ICST, pp. 1–6, 2016, doi: 10.1109/ICSensT.2016.7796258.

[207] G. Gioberto and L. E. Dunne, “Overlock-stitched stretch sensors: Characterization and effect of fabric property,” J. Text. Apparel, Technol. Manag., vol. 8, no. 3, pp. 1–14, 2013.

[208] A. M. Pisa, G. Chernyshov, A. F. Nassou, and K. Kunze, “Towards interactive mindfulness training using breathing based feedback,” UbiComp/ISWC 2017 - Adjun. Proc. 2017 ACM Int. Jt. Conf. Pervasive Ubiquitous Comput. Proc. 2017 ACM Int. Symp. Wearable Comput., pp. 688–692, 2017, doi: 10.1145/3123024.3129268.

[209] C. D. Spielberger, R. L. Gorsuch, R. Lushene, P. R. Vagg, and G. A. Jacobs, “State-Trait- Anxiety-Inventory for Adults Self-Evaluation Questionnaire STAI Form Y-1 nad Form Y- 2.” p. 5, 1968.

[210] S. A. Jackson and H. W. Marsh, “Development and validation of a scale to measure optimal experience: The flow state scale,” J. Sport Exerc. Psychol., vol. 18, no. 1, pp. 17–35, 1996, doi: 10.1123/jsep.18.1.17.

[211] T. Brandmeyer and A. Delorme, “Reduced mind wandering in experienced meditators and associated EEG correlates,” Exp. Brain Res., vol. 236, no. 9, pp. 2519–2528, 2018, doi: 10.1007/s00221-016-4811-5.

[212] D. MacLean, A. Roseway, and M. Czerwinski, “MoodWings: A wearable biofeedback device for real-time stress intervention,” ACM Int. Conf. Proceeding Ser., 2013, doi: 10.1145/2504335.2504406.

[213] L. M. Braunstein, J. J. Gross, and K. N. Ochsner, “Explicit and implicit emotion regulation: A multi-level framework,” Soc. Cogn. Affect. Neurosci., vol. 12, no. 10, pp. 1545–1557, 2017, doi: 10.1093/scan/nsx096.

[214] A. Gyurak, J. J. Gross, and A. Etkin, “Explicit and implicit emotion regulation: A dual- process framework,” Cogn. Emot., vol. 25, no. 3, pp. 400–412, 2011, doi: 10.1080/02699931.2010.544160.

[215] Y. Y. Tang, B. K. Hölzel, and M. I. Posner, “The neuroscience of mindfulness meditation,” Nat. Rev. Neurosci., vol. 16, no. 4, pp. 213–225, 2015, doi: 10.1038/nrn3916.

184

[216] E. Zurich, “Sensory Motor Systems.” [Online]. Available: https://www.iris.ethz.ch/the- institute/sensory-motor-systems-lab.html.

[217] E. Ayvali and J. Desai, “Pulse Width Modulation-based Temperature tracking for Feedback Control of Shape Memory Alloy Actuator,” J Intell Mater. Sys Struct, vol. 25, no. 6, pp. 720–730, 2014, doi: 10.1038/jid.2014.371.

[218] N. Ma and G. Gong, “Control of shape memory alloy actuator using pulse width modulation,” Smart Mater. Struct., vol. 12, pp. 712–719, 2003.

[219] J. P. Swensen and A. M. Dollar, “Optimization of parallel spring antagonists for Nitinol shape memory alloy actuators,” Proc. - IEEE Int. Conf. Robot. Autom., pp. 6345–6349, 2014, doi: 10.1109/ICRA.2014.6907795.

[220] M. Priebe, E. Foo, and B. Holschuh, “Shape Memory Alloy Haptic Compression Garment for Media Augmentation in Virtual Reality Environment,” UIST 2020- ACM Symp. User Interface Softw. Technol., pp. 15–17, 2020, doi: https://doi.org/10.1145/3313831.

[221] R. Ashford, “An exploration of responsive and emotive wearables through research prototyping,” Conf. Hum. Factors Comput. Syst. - Proc., no. 2013, pp. 2013–2016, 2019, doi: 10.1145/3290607.3313284.

[222] G. Mura, “Wearable technologies for emotion communication,” Metu J. Fac. Archit., vol. 25, no. 1, pp. 153–161, 2008.

[223] O. Tomico and D. Wilde, “Soft, embodied, situated & connected: enriching interactions with soft wearables,” mUX J. Mob. User Exp., vol. 5, no. 1, pp. 1–17, 2016, doi: 10.1186/s13678-016-0006-z.

[224] O. Tomico, L. Hallnäs, R. H. Liang, and S. A. G. Wensveen, “Towards a next wave of wearable and fashionable interactions,” Int. J. Des., vol. 11, no. 3, pp. 1–6, 2017.

[225] F. H. Wilhelm et al., “Construction and Validation of the Touch Experiences and Attitudes Questionnaire (TEAQ): A Self-report Measure to Determine Attitudes Toward and Experiences of Positive Touch,” Biol. Psychol., vol. 42, no. 3, pp. 379–416, 2001, doi: 10.1016/S0301-0511(01)00113-2.

185

[226] L. E. Osborn, “Reflex: A Closed-Loop Tactile Feedback System for Use in Upper Limb Prosthesis Grip Control,” 2014.

[227] B. M. Kudielka, H. Hellhammer, and C. Kirschbaum, “Ten years of research with the trier social stress test.,” Soc. Neurosci., no. January 2007, pp. 56–83, 2007.

[228] F. X. Lesage, S. Berjot, and F. Deschamps, “Clinical stress assessment using a visual analogue scale,” Occup. Med. (Chic. Ill)., vol. 62, no. 8, pp. 600–605, 2012, doi: 10.1093/occmed/kqs140.

[229] C. Kirschbaum, K. M. Pirke, and D. H. Hellhammer, “The ‘Trier social stress test’ - A tool for investigating psychobiological stress responses in a laboratory setting,” Neuropsychobiology, vol. 28, no. 1–2, pp. 76–81, 1993, doi: 10.1159/000119004.

[230] M. Kotlyar et al., “Timing of nicotine lozenge administration to minimize trigger induced craving and withdrawal symptoms,” Addict. Behav., vol. 71, pp. 18–24, 2017, doi: 10.1016/j.addbeh.2017.02.018.

[231] M. Kotlyar et al., “Effect of paroxetine on physiological response to stress and smoking,” Psychosom. Med., vol. 75, no. 3, pp. 236–243, 2013, doi: 10.1097/PSY.0b013e3182898f6d.

186

Appendices

High-fidelity Garment Perceived Intensity Statistical Results Summary Table

The following section (Table A-1) details summary statistics regarding differences between perceived intensities (of low, medium, high), for various locations on the body, given the high- fidelity garment actuation. The presented tables are supplemental to the results presented in Chapter 4, Section 4.2 (Figure 4-3, Figure 4-4, Figure 4-5). The non-parametric Kruskal-Wallis rank sum test was employed, and significant differences observed are as indicated.

Table A-1: Perceived intensity Kruskal Wallis Rank Sum Test summary table, divided into three location/ compression vector blocks (i) Torso and Arms, (ii) Shoulder-straight conditions, and (iii) Shoulder- diagonal conditions. Perceived Intensity Kruskal Wallis Rank Sum Test Summary Table Torso and Arms (Supplemental to Figure 4-3) Body Location χ2 Df p-value Signif. codes Sides 0.01597 2 0.992 Abdomen 0.11189 2 0.9456 Lower Back 0.62339 2 0.7322 Arms 5.2954 2 0.0708 . Shoulder-Straight (Supplemental to Figure 4-4) Body Location χ2 Df p-value Signif. codes Front Side Chest 1.2262 2 0.5417 Front Mid Chest 0.89346 2 0.6397 Top of Shoulders 4.9823 2 0.0828 . Back of Neck 5.7909 2 0.0553 . Upper Back 0.74599 2 0.6887 Shoulder-Diagonal (Supplemental to Figure 4-5) Body Location χ2 Df p-value Signif. codes Front Side Chest 0.074236 2 0.9636 Front Mid Chest 0.28674 2 0.8664 Top of Shoulders 6.939 2 0.0311 * Back of Neck 4.3325 2 0.1146 Upper Back 0.69103 2 0.7079 Signif. codes: 0 ‘***’, 0.01 ‘**’, 0.05 ‘*’, 0.1 ‘.’

187

High-fidelity Garment CALM Pressure Comfort Statistical Results Summary Table

The following section (Table B-2) details summary ANOVA statistical results of varying comfort levels given the high-fidelity garment actuation, comparing between conditions for each body location. This is supplemental to the results presented in Chapter 4, Section 4.2 (Figure 4-6 and Figure 4-7).

Table B-2: CALM Pressure comfort ANOVA summary table, comparing conditions between various body locations CALM Pressure Comfort ANOVA Summary Table Body Location Df Sum Sq Mean Sq F value p-value Signif. codes Trunk of Body group 4 6000 1500 1.91 0.117 residuals 79 62049 785.4 Abdomen group 4 5808 1452.1 1.859 0.126 residuals 79 61714 781.2 Middle/Lower Back group 4 1285 321.3 0.446 0.775 residuals 79 56859 719.7 Upper Arms group 4 22942 5736 6.041 3.00E-04 *** residuals 72 68356 949 Front of Chest group 7 10834 1548 1.514 0.168 residuals 127 129821 1022 Top of Shoulders group 7 13237 1891 2.237 0.0353 * residuals 127 107358 845.3 Upper Back group 7 10134 1447.7 1.597 0.142 residuals 127 115139 906.6 Signif. codes: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’

For the locations which the ANOVA analyses revealed significant differences, the next step is to perform follow up tests to determine which test combinations groups were different. The following table (Table B-3 and Table B-4) provides the summary statistics for the post hoc analyses with the Tukey multiple comparison test to reveal comfort rating differences between groups, specifically for the upper arms and shoulder regions.

188

Table B-3: Tukey multiple comparisons of means summary table for the upper arm. Tukey Multiple Comparisons of Means for the Upper Arms diff lwr upr p adj Signif. codes Arm.Dur-Arm.Base -8.961 -41.000 23.076 0.934 Arm.High-Arm.Base -42.033 -73.019 -11.047 0.002 ** Arm.Low-Arm.Base -21.470 -52.4565 9.514 0.306 Arm Med-Arm.Base -43.095 -74.081 -12.110 0.001 ** Arm.High-Arm.Dur -33.071 -64.623 -1.519 0.035 * Arm.Low-Arm.Dur -12.508 -44.060 19.042 0.801 Arm.Med-Arm.Dur -34.133 -65.685 -2.582 0.027 * Arm.Low-Arm.High 20.562 -9.919 51.044 0.333 Arm.Med-Arm.High -1.062 -31.544 29.419 0.999 Arm.Med-Arm.Low -21.625 -52.106 8.856 0.283 Signif. codes: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’

Table B-4: Tukey multiple comparisons of means summary table for the shoulders Tukey Multiple Comparisons of Means for the Shoulders diff lwr upr p adj Signif. codes Shld.Diaghigh-Shld.Base -22.058 -52.795 8.677 0.351 Shld.Diaglow-Shld.Base -16.176 -46.912 14.559 0.736 Shld.Diagmed-Shld.Base -21.705 -52.442 9.030 0.372 Shld.Dur-Shld.Base -1.397 -32.610 29.815 0.999 Shld.Strhigh-Shld.Base -31.823 -62.559 1.087 0.036 * Shld.Strlow-Shld.Base -15.176 -45.912 15.559 0.794 Shld.Strmed-Shld.Base -17.352 -48.089 13.383 0.661 Shld.Diaglow-Shld.Diaghigh 5.882 -24.854 36.618 0.998 Shld.Diagmed-Shld.Diaghigh 0.352 -30.383 31.089 1.000 Shld.Dur-Shld.Diaghigh 20.661 -10.551 51.874 0.459 Shld.Strhigh-Shld.Diaghigh -9.764 -40.501 20.971 0.976 Shld.Strlow-Shld.Diaghigh 6.882 -23.854 37.618 0.999 Shld.Strmed-Shld.Diaghigh 4.7058 -26.030 35.442 0.999 Shld.Diagmed-Shld.Diaglow -5.529 -36.265 25.207 0.999 Shld.Dur-Shld.Diaglow 14.779 -16.433 45.992 0.827 Shld.Strhigh-Shld.Diaglow -15.647 -46.383 15.0893 0.767 Shld.Strlow-Shld.Diaglow 1.000 -29.736 31.736 1.000 Shld.Strmed-Shld.Diaglow -1.176 -31.912 29.559 1.000 Shld.Dur-Shld.Diagmed 20.308 -10.904 51.521 0.482 Shld.Strhigh-Shld.Diagmed -10.117 -40.854 20.618 0.971 Shld.Strlow-Shld.Diagmed 6.529 -24.207 37.265 0.997 Shld.Strmed-Shld.Diagmed 4.352 -26.383 35.089 0.999 Shld.Strhigh-Shld.Dur -30.4264 -61.639 0.786 0.061 . Shld.Strlow-Shld.Dur -13.779 -44.992 17.433 0.873 Shld.Strmed-Shld.Dur -15.955 -47.168 15.257 0.763 Shld.Strlow-Shld.Strhigh 16.647 -14.089 47.3835 0.706 Shld.Strmed-Shld.Strhigh 14.470 -16.265 45.207 0.831 Shld.Strmed-Shld.Strlow -2.176 -32.912 28.5599 0.999 Signif. codes: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’

189

Emotion Communication Survey Study Participants’ Demographic Breakdown

For the survey study on ‘warm touch’ haptic communication conducted in Chapter 5, since we know that perceptions towards touch communication can be influenced by cultural factors, we gathered some basic information on our participants’ demographics, even though the survey was deployed for participants residing in the United States (self-reported or as registered through Amazon Mechanical Turk). Table C-5 presents the country of origin breakdown of each of our survey participants; Table C-6 presents information on the number of years the participants have spent in the United States of America for each of the survey.

Table C-5: Country of origin information for participants of both surveys Country of Origin Survey 1 Survey 2: Communicate Survey 2: Elicit United States of America 46 40 40 India 1 - - Malaysia 3 - - Islamic Republic of Iran 1 - - China 2 - - Total 53 40 40

Table C-6: Number of years participants from both surveys have spent in the United States Number of Years Spent in U.S. Survey 1 Survey 2: Communicate Survey 2: Elicit 1-5 years 4 - - 6-10 years 1 - - 11-20 years 2 - - More than 20 years 46 40 40 Total 53 40 40

190

Emotion Communication Survey 1 Study’s Parallel Sets Plot Data Visualization

In Chapter 5, one major contribution from the emotion communication survey was to determine ‘warm touch’ haptic strategies employed by participants during touch communication with a friend. The relationship between ‘warm touch’ parameter dimensions for each emotion were visualized as parallel set plots; only representative plots were presented in the main text due to the excessive number of plots (refer to Chapter 5, Section 5.2.2). The following section presents parallel set plots for all emotions: Figure D-2 presents the visuals for Ekman’s basic emotions including Sadness, Happiness, Fear, and Anger. Figure D-3 presents the plots for prosocial emotions such as Love and Gratitude, and other emotions including Calm and Attention.

Figure D-2: Survey 1’s parallel sets plots data visualization for Ekman’s basic emotions. (a) Row 1 left: Sadness, (b) Row 1 right: Happiness, (c) Row 2 left: Fear, (d) Row 2 right: Anger

191

192

Emotion Communication Survey 1’s Pairwise Comparisons of Confidence Ratings between Emotions

For Survey 1 conducted in Chapter 5, we also gathered participants’ self-reported confidence for their selection of the ‘warm touch’ haptic strategies in communicating the targeted emotion. Kruskal-Wallis test on participants’ confidence for ‘warm touch’ selections showed statistically significant differences between emotions, hence post-hoc pairwise comparisons with Wilcoxon rank sum test with Bonferroni corrected p-values were conducted. The summary statistics table is presented in Table E-7 (supplementing the information presented in Chapter 5, Section 5.2.2).

Table E-7: Pairwise comparisons of confidence ratings using paired Wilcoxon rank sum test with Bonferroni corrected p-values. anger attention calm fear gratitude happiness love attention 1.0000 ------calm 1.0000 1.0000 - - - - - fear 1.0000 1.0000 1.0000 - - - - gratitude 1.0000 0.2727 1.0000 0.1700 - - - happiness 1.0000 1.0000 1.0000 0.9847 1.0000 - - love 1.0000 1.0000 1.0000 1.0000 0.8220 1.0000 - sadness 0.1426 0.0039* 0.0179* 0.0024* 0.2585 0.5568 0.0229* *An asterisk indicates a significant result (p < 0.05).

193

Emotion Communication Survey 2 Study’s Parallel Sets Plot Data Visualization

Similarly, in Chapter 5, another contribution from the emotion communication survey was to tease out potential differences in ‘warm touch’ haptic strategies employed by participants if the task was framed as a ‘communicate’ vs. ‘elicit’ scenario. As before, the relationship between ‘warm touch’ parameter dimensions for each emotion were visualized as parallel set plots. In the main text, only plots for emotions where significant differences were observed were presented (refer to Chapter 5, Section 5.3.2). The following section presents parallel set plots for all remaining emotions: Figure F-4 presents the visuals for Ekman’s basic emotions including Sadness, Happiness, Fear, and Anger. Figure F-5 presents the plots for prosocial emotions such as Love and Gratitude, and other emotions including Calm. Note that for these figures, the first (left) column presents results for Group 1 (Communicate), while the second column involves results obtained from Group 2 (Elicit).

Group 1: Communicate Group 2: Elicit

194

Figure F-4: Survey 2’s quantitative survey data visualization for Ekman’s basic emotions First column: Group 1 (Communicate); Second column: Group 2 (Elicit)

195

Group 1: Communicate Group 2: Elicit

Figure F-5: Survey 2’s quantitative survey data visualization for prosocial and other emotions. First column: Group 1 (Communicate); Second column: Group 2 (Elicit)

196

Emotion Communication Study Proposal: Encoding-Decoding Communication Task Mediated by SMA Compression Garment

As mentioned in Chapter 5 Section 5.4.1, while the online surveys deployed to understand user expectations and strategies allowed us the gathering of information from a larger, general population, the haptic communication itself remains hypothetical—the results tells us what users expect to perform—which could vary when faced with an actual situation. There is a possibility that with the use of an actual garment with experienced haptics, the ‘warm touch’ strategies and mental models might vary. The initial intent was to execute the study with the use of an actual garment with haptics that users can experience. However, due to restrictions imposed by COVID- 19, the study has been postponed indefinitely. The outline of the proposed study, along with the a designed garment in Figure G-6 (based on prior feedback from Chapters 3 to 6), as well as the study’s facilitator script is presented below.

Key Idea: Users convey and identify specified emotions with a partner using garment-driven haptics alone (without audio or visual support interaction—to see what could be conveyed by compression and warmth haptics in isolation).

Method Overview: - Dyads to participate in each session; gender balanced/cross-gendered pairs, one male one female. - Participants are randomly assigned to initial roles of ‘sender’ and ‘receiver’ and performed the affective touch communication task. The touch communication task involves the ‘sender’ designing a series of touch features corresponding to a specific emotion (prompted with emotion word and scenario example), given varying controllable parameters, on a computer interface. - The designed haptic features are then delivered through a computer-mediated channel to haptic garments worn by the receiver. - The ‘receiver’ will then attempt to indicate what emotion they thought the ‘sender’ was communicating. - The entire task consists of 8 emotions (presented in random order), including 4 of Ekman’s basic emotions (Sadness, Fear, Happiness, Anger), 2 prosocial emotions (Love, Gratitude), and two functional (Calm, Attention: as reference). - After performing the task, participants switch roles so that the ‘sender’ becomes the ‘receiver’ and vice versa. - At the end of the session, both participants were asked to join in on a qualitative interview.

Participant selection: - Men and women self-identified as physically healthy (no cardiovascular issues, not 197

pregnant, have not had surgery in the past 6 months), wearing size M, fluent in English, ages 18-34. - Participants are strangers and have not been acquainted prior to the study itself. Target participant pool of 12 pairs, n=24. - Participant demographics and information will be collected, including but not limited to the country of origin and years spent in the United States (potential cultural associations to explain potential outliers), as well as attitudes towards technology, experience with wearable technology, perceptions towards social touch (i.e., touch aversion or inclination).

Haptic devices: The following variables can be modulated by the ‘sender’: (1) Body location: Shoulders, torso, neck, upper arms, hand, wrist (all NVBP, one VBP on torso) (2) Pattern: Constant (binary on/off and allowed to control duration) vs. pulsing (set rate of perceptible pulse) (3) Intensity: Low vs. High

A prototype of the system designed using feedback from Chapters 3-6 is presented in Figure G-6 [220]. The system includes SMA actuators capable of presenting warm, compressive forces on the torso, shoulders, arms, hands, and neck (neck piece not featured in the figure) at 2 different intensities (low and high) and varying stimuli patterns (constant vs. pulsing).

Measures: (1) Haptic design of touch cues by the ‘sender’. This will reveal objective strategies used to represent various touch cues. Clustering analysis can be used to unveil strategies that may lead to more successful emotional touch communication and may help in specifying future design requirements for such haptic actuators, or the types of emotions that cannot be represented well 198

with such haptic actuators. Possibly be presented as body maps for men→ women and women→ men touch communication.

(2) Modified forced-choice response of emotional label. After the ‘sender’ signaled that they had finished ‘sending’ the touch communication, ‘receivers’ will be presented with buttons on a screen (survey) labeled with receiver choices and must select what they thought the ‘sender’ was communicating. If receiver selected ‘other’, they will be prompted to type a response. This performance measure will result in a confusion matrix and can be compared to chance rate of 12.5%.

(3) Open ended response in support of reason for presenting (the ‘sender’) and choosing (the ‘receiver’) the haptic strategy or response to represent the emotion label. After the cue is delivered, the ‘sender’ will be asked why they chose those specific parameters to represent the emotion label. For the ‘receiver’, after selecting the force-choice emotional label, they will be asked to explain what cues they used to make the decision on the emotion label. This will help explain the thought process as to what mental model participants use to map emotional meanings to the haptic sensations, or the type of metaphoric associations they might use.

(4) Confidence score for both ‘sender’ and ‘receiver’ will give us insight into the ease of conveying and perceiving specific emotions. This self-reported rating can be correlated with performance measures to see if there are emotions that are easier to present and convey with this specific set of haptic stimuli.

(5) Social Touch Questionnaire (STQ) [225] will be presented at the beginning to assess personal attitudes towards social situations involving touch. This information will supplement the cultural background information and can be correlated with touch communication performance measures, as well as the difference between STQ measures of the sender and receiver and how that may impact performance.

(6) Wanting of touch communication. At the end of each communication task (before switching roles or the interview), participants were asked ‘In general, how much do you want to communicate [cue] to other people through touch?’ (sender) and ‘In general, how much do you want other to communicate [cue] to you through touch?’ (receiver). Answers were marked across a visual analogue scale with ‘not at all’ and ‘very much’ marked at the ends of the line.

(7) Empatica E4 on ‘receiver’ to collect physiological signals. The device will be able to collect participants’ EDA, HR, and BVP.

(8) Paired semi-structured qualitative interview will be performed together with both participants. Participants will be provided with pen and paper to sketch and design together. Questions will be centered on user expectations, wants of touch communication, hedonic value of computer- mediated touch communication, pragmatic value of computer-mediated touch communication, things user might want to change, situations and ideas on how to use the devices. 199

Study Script: Welcome Briefing + Consent Process Estimated Time: 20 minutes • Ushering and consent process to be completed in two separate rooms; both participants will be briefed separately using a standardized script. • ‘Receiver’ will be welcomed in the main study room, seated as a spot with obstructed view from door, i.e., behind a curtain. • ‘Sender’ will be welcomed at WTL and started with consent process.

(Start process with ‘receiver’ at the main study room)

“Thank you for agreeing to take part in our study. My name is (; or in the case of ‘sender’ participant, ). I will be working from a script to ensure that my instructions to every participant in the study are the same.”

“We are here to learn about the communication of emotions using only the sense of touch, mediated by various haptic-based wearable technologies.”

“We have a consent form that we need you to read carefully. We will walk you through the document by providing you a brief overview of the study. Please do not sign the last page until we tell you so.” (Hand consent form to participant)

“In this experiment, you will have the opportunity to design and execute the communication of various types of emotions to another person, whom you have not met, using only computer- mediated touch signals. The computer-mediated haptics will be delivered through a set of wearable technology garments that apply compression and warmth to stimulate a touch.”

“The actuators to simulate these touch sensations are made of shape memory alloys; they are smart metals that can ‘remember’ its original shape after deformation when it is heated. The heating method we will be using is through resistive heating, like how a wire heats up when current flows through it. However, do not worry, we have taken care to insulate all the actuators and wires that we will be using in the garment and the current we will be applying will be nowhere near harmful. However, if you feel that the garment becomes too tight or too warm, please let us know immediately and we will fix the problem or terminate the test.”

“To give you an overview of what to expect in this session, first, you will have the opportunity to familiarize yourself with the wearable technology garments and their various parameter controls, including being able to control and feel the different touch intensities, location, duration, and patterns. Using these parameters, you will be asked to design strategies to communicate various types of emotions. You will be given a list of 8 emotion words to design for and you will be given approximately 8 minutes to familiarize yourself with the controls, sensations, and design the haptic strategies.”

“The strategies that you have designed will then be sent to another participant through a similar set 200 of wearable technology garments, and he/she will be asked to guess the emotion that you are trying to communicate. Upon the completion of all emotion words, your roles will then be swapped. The order to which you are either the ‘sender’ or ‘receiver’ will be randomized.”

“Throughout the study, you may be audio/video recorded. However, your name will not be attached to any of that data and we only employ secure university cloud storage methods. Once the data is analyzed and transcribed, those audio and video files will be immediately deleted.”

“You will also be asked to answer some survey questions throughout the process, and we will end the study with some interview about your experience. Again, your interview responses will be recorded and once your responses are transcribed, the audio files will be immediately deleted. I will also be taking notes throughout the study.”

“However, since we are trying to investigate the communication of emotions through a purely haptic channel, I ask that you remain as silent as possible once we start the study, and only speak up during emergencies or signaling your intent to discontinue the study. Should you have other non-urgent questions including the operation of the garment, we will be able to communicate through written text.”

“This session is estimated to take approximately 2.5 hours. At any point of the test, should you need to stop, please inform us immediately. Should garment become too uncomfortable, please let us know immediately and we will fix the problem or terminate the test. Should you choose to discontinue the study after the experiment begins, your compensation and your relationship with the university will not be affected.”

“To assess your understanding of what you just heard and read, in your own words, please briefly explain what you think is going to happen in the study.” (Check for understanding)

“Thank you, you can now sign both of the consent forms. One of the copies is for you to keep.” (Collect consent form and leave one for participant)

“Do you have any questions before we begin?” (Answer any questions that the participant may have)

Baseline Survey + Ushering Estimated Time: 5 minutes • Demographic information (age, gender, country of origin, years spent in the US) • Social Touch Questionnaire (STQ) For ‘receiver’ participant: “Now, I will need to attend to the other participant. I will be back shortly. In the meantime, please fill up this baseline survey.” (Hand participant the baseline survey) (Leave room to receive ‘sender’ participant)

(In other room/WTL where the other participant is going through the consent process or waiting) 201

For ‘sender’ participant: “Hi, I am and will be running the rest of the study with you today. I was going through the same consent process with the other participant in another room, thank you for your patience and I am excited to have you join us in this study.”

“Now, we will head to the room where the other participant is located. Again, as a reminder, please try to remain silent once we enter the room and start the study.” (Head to main study room) (Instruct participant to sit facing each other, separated by curtain)

(Moderator seated at the head of table, in between the curtain so that both participants are visible.)

Garment Fitting + Individual Design Task Estimated Time: 15 minutes • Subjects will not be allowed to discuss strategies with one another; they will be instructed to remain silent throughout this process unless they have system operational questions.

“Now, I will help you put on these wearable technology garments.” (Help each participant put on the garment)

“This is the design phase of the study where you will have the chance to experience the different touch intensities, location, duration, and patterns as you vary the garment controls using the interface on the laptop in front of you. The interface is currently connected to the garment you are wearing right now so as you change the parameters, you should feel the different touch sensations generated by the garment, on your body.”

“You will also notice an envelope on the table. Please open the envelope and take out the sheet of paper. On that paper, you should see a list of 10 emotion words, along with a prompt to describe each emotion cue. Your task is to design ways to use the wearable garment’s touch sensations to communicate and represent that specific emotion.” (Make sure participants have access to envelope with emotion cue list.)

“Please indicate on the paper on your design strategies as you feel the garment haptics and think through how you want to represent each emotion.”

“You will be given approximately 8 minutes to familiarize yourself with the controls, sensations, and design the haptic strategies.”

“You also perhaps have noticed that there is someone else sitting across you, separated by this curtain. Later in the study, whatever strategy you have designed will be sent to your study partner and they will be asked to guess the emotion you are trying to communicate. The initial order of who will be the ‘sender’ and ‘receiver’ will be revealed after this design task. Both of you will have the opportunity to send and receive the touch signals, meaning you will swap roles after each of the list is completed.”

“Before you begin, please put on the noise-cancelling headphones in front of you, this is to make sure we don’t have audio interference and you can focus on the task at hand. Just speak up if you 202 have questions about the controls and the interface and I will help you with it. I will provide you with a signal once the time is up.” (Make sure participants have their headphones on.) (Set timer for 8 minutes.) (Help participants with the control interface if necessary.)

Role Assignment + Establish connection to partner’s garment Estimated Time: 10 minutes (Signal to participants 8 minutes is up and to remove their headphones.) “I hope you had time to work through your haptic design strategies. Now as mentioned, one of you will be the ‘sender’ where you will send your designed haptics to your study partner and the other will be the ‘receiver’ where you will make a guess on what emotion your study partner is trying to communicate. Your roles will be swapped after the list is complete.”

“You will begin as the ‘sender’ for this first part of the study.” (Assign role to participant.)

“And you will begin as the ‘receiver’ for this first part of the study.” (Assign role to participant.)

To ‘receiver’ participant: “So first you will put on this wrist-band device, called an Empatica E4. This device is like an apple watch that collected physiological signals such as heart rate, electrodermal activity etc. This will allow us to monitor if there are any physiological changes associated with the user of the wearable garments. Between each emotional cue, I will need you to press this event mark button so we can distinguish the periods between each emotional cue for data analysis. Your study partner will be sending the designed haptic sensations to your garment and your task is to figure out what emotion they are trying to communicate. This light will be on during your study partner’s sending task, once they are done, the light will turn off and you can proceed to making your guess on this laptop survey.”

“Now you can put your headphones on and wait for the LED signal to turn on and receive the haptic sensation.”

To ‘sender’ participant: “Now I will help you connect your controls to your study partner’s garment.” (Change the Bluetooth connection) “So, your task now is to communicate the emotions using your designed haptic sensations, in the order of emotional prompt presentations shown on the screen. There will also be a short survey prompt in between each emotional cue.”

203

“Since your study partner is now wearing the noise-cancelling headphones, please feel free to communicate with me if you have any questions.” (Answer any lingering questions) “You may now begin.”

Emotional Touch Communication 1 + Survey Estimated Time: 40 minutes • For the ‘sender’, they will be presented with a randomized emotion word cue and scenario. Their role is to present the designed touch features corresponding to that emotion cue, given varying controllable parameters through a computer interface. • The designed haptic features are then delivered through a computer-mediated channel to haptic garments worn by the receiver. • The ‘receiver’ will then attempt to indicate what emotion they thought the ‘sender’ was communicating, the emotional cue modified forced-choice response order in the survey will be randomized, with an ‘other’ open-ended option.

“Please communicate this emotion.” (Present emotion cue card with scenario—to be repeated for all emotions) (Researcher to manually provide randomized emotion cue card to ‘sender’) (The survey will lead most of the study for questionnaires in between each emotional cue.)

• The ‘sender’ and ‘receiver’ will be asked to rate their confidence (Likert scale) and reasons for designing/interpreting the emotion and sensations (open-ended response). • Upon completion of all 8 emotion word pair, the participants will answer a short survey on the wanting of touch communication from both the ‘sender’ and ‘receiver’ point of view.

(The actions of the ‘sender’ will be recorded, i.e., parameters the ‘sender’ is using for haptic design.) (Remember to turn on/off the LED signaling light for the ‘receiver’) (Observe and note any changes to receiver facial expressions)

Swap devices + Establish connection to partner’s garment Estimated Time: 5 minutes (Signal participants to remove their headphones) “Thank you for completing the first part of the study. Now, your roles of ‘sender’ and ‘receiver’ will be swapped.” (Remove Empatica E4 from initial ‘receiver’ participant)

To new ‘receiver’ participant: “Now that you are the ‘receiver’, you will wear this Empatica E4 to collect physiological signals such as heart rate, electrodermal activity etc. Between each emotional cue, I will need you to press this event mark button so we can distinguish the periods between each emotional cue for data analysis. Your study partner will be sending the designed haptic sensations to your garment and your task is to figure out what emotion they are trying to communicate. This light will be on during your study partner’s sending task, once they are done, the light will turn off 204 and you can proceed to making your guess on this laptop survey.”

“Now you can put your headphones on and wait for the LED signal to turn on and receive the haptic sensation.”

To new ‘sender’ participant: “Now I will help you connect your controls to your study partner’s garment, and you can review your design planning sheet in the meantime.” (Change the Bluetooth connection) “So, your task now is to communicate the emotions using your designed haptic sensations, in the order of emotional prompt presentations shown on the screen. There will also be a short survey prompt in between each emotional cue.”

“After reading each prompt, turn on this LED light to signal to your study partner you are going to send the haptics to their garment and turn it off when you are done with sending your design.” (Make sure to turn on the LED light when ‘sender’ is sending the cue to ‘receiver’ participant.) “Since your study partner is now wearing the noise-cancelling headphones, please feel free to communicate with me if you have any questions.” (Answer any lingering questions) “You may now begin.”

Emotional Touch Communication 2 + Survey Estimated Time: 40 minutes • For the ‘sender’, they will be presented with a randomized emotion word cue and scenario. Their role is to present the designed touch features corresponding to that emotion cue, given varying controllable parameters through a computer interface. • The designed haptic features are then delivered through a computer-mediated channel to haptic garments worn by the receiver. • The ‘receiver’ will then attempt to indicate what emotion they thought the ‘sender’ was communicating, the emotional cue modified forced-choice response order in the survey will be randomized, with an ‘other’ open-ended option.

• The ‘sender’ and ‘receiver’ will be asked to rate their confidence (Likert scale) and reasons for interpreting the emotion and sensations (open-ended response). • Upon completion of all 8 emotion word pair, the participants will answer a short survey on the wanting of touch communication from both the ‘sender’ and ‘receiver’ point of view. (The actions of the ‘sender’ will be recorded, i.e., parameters the ‘sender’ is using for haptic design.) (Remember to turn on/off the LED signaling light for the ‘receiver’) (Observe and note any changes to receiver facial expressions.) 205

Paired Qualitative Interview Estimated Time: 20 minutes • After getting out of the haptic garments, both participants are now invited to participate in a qualitative, semi-structured interview together.

(Signal participants to remove their headphones.) “Thank you for completing the tasks. The last part of this study is a qualitative interview, and you will both participate in this section together.” (Remove curtain partition to reveal study partner.)

“Please know that we are not testing you and there is no such thing as a wrong answer. We would like to know exactly what you think, not what you think I want to hear. Being honest in your feedback will really help us in understanding what works and what doesn’t work.” (Lay out craft items including colored pens and papers)

“At this time, I encourage you to openly share your experiences and even work with your partner to answer some of these questions. You see a variety of craft items in front of you now. Feel free to use those tools to illustrate your points as we go through this interview. For instance, you can sketch out scenarios using stick figures or illustrate areas of the body in relation to describing the design of haptic sensations.”

(Semi-structured survey, follow-up questions and probes included in each question, i.e. for multi- questions listed below, they will not be presented all-together)

1. How did the experience align or conflict with your expectation of the study with haptic- based emotional communication? How easy or difficult was it?

2. Could you both elaborate more on the emotional cues you want to communicate the most through touch? Was that the same of different for receiving the touch communication? Why is that so?

3. What do you think makes it easier or more difficult to convey and perceive emotions with this method? Are there anything you’d like to change or ways you think will make your performance better? (Frame this: The reason we’re asking this is because as you can imagine, there are a lot of ways to approach this topic, but to make this garment, we had to make some decisions. So we’d like to know if we can decide differently, or would you recommend other types of stimuli for certain emotions?)

4. How much did you enjoy the computer-mediated haptic sensations? What did you like and dislike about them? What did you think about the temperature and compression stimuli?

5. What did you like/dislike on the comfort of the garment?

206

6. Can you share what you think about the benefits and problems of using this type of communication devices might be?

7. In this study, you both are not acquainted and trying to communicate with each other. If this was a different situation where you knew the person you are communicating with, for instance a friend or a loved one, how would your attitudes about your experience change? Are there specific use scenarios you think this technology might be helpful? Feel free to think big and craft out specific situations—you can even sketch out what you think.

8. Thinking beyond just this experiment you just experienced, what other applications do you think you could or want to use these devices for?

Debriefing + Gift Card Estimated Time: 2 minutes “That’s all that we have. Are the questions about the study that I can answer for you before we end?” (Answer questions)

“Thank you again for participating in our study, here is a gift card as a note of thanks for your help and time. (Provide participants with gift cards)

207

Haptic-Augmented Meditation Study’s Force Sensor Calibration and Measurement

Figure H-7 shows the fabric-based force sensors manufactured (according to [226]) that were used to estimate pressures provided by the compression garment.

Figure H-7: Fabric-based force sensors (two conductive materials sandwiching Eeyonyx material) manufactured according to [226]

Measuring absolute pressures for on-body applications are particularly challenging since the human body is soft and contoured; this is especially so for compressive forces as in this situation (as the normal and shear forces are so tightly intertwined). To account for the soft, contoured surfaces, the use of sensors to estimate on-body pressures as similar as possible to the conditions in which they will be used in measurement is necessary. Hence, the sensors were calibrated according to an adapted method suggested by [169], whereby the sensor was placed over a polyurethane foam- covered glass cylinder (circumference= 30cm) and a blood pressure cuff was placed over the sensor. The cuff was then inflated from 20mmHg to 60mmHg and analog voltage output collected by the NI-6001 data acquisition system was recorded (60mmHg was set as the upper limit due to sensor saturation and the forces provided by the garment are not expected to exceed those of medical grade compression garments ~50mmHg—in fact, for the purposes of the meditation study, we were targeting a low but detectable compression sensations).

208

Figure H-8: Force Sensor Calibration Curve

The collected data was fitted to a 2-point power law (Figure H-8) and the equation obtained was used to calculate the estimated pressures encountered by participants. Table H-8 provides the maximum and minimum analog voltage readings for each participant. The average maximum and minimum recorded voltages were calculated and substituted into the power law equation obtained from the calibration process. This results in estimated averaged pressures of 15-24mmHg provided by the SMA-garment capable of providing compressive forces to users.

Table H-8: Analog voltage readings collected by Data Acquisition System that are used to calculate estimated pressures provided by the garment Participant # Maximum Recorded Voltage Minimum Recorded Voltage 1 4.20 4.39 2 4.43 4.52 3 3.86 4.06 4 4.16 4.33 5 4.22 4.29 6 4.32 4.44 7 3.41 3.64 8 4.05 4.31 9 4.2 4.32 10 3.48 3.81 Average 4.033 4.211 Corresponding Estimated 15.21 23.64 Pressures (mmHg)

209

Emotion Regulation Study Proposal: Stressor Task SMA-based Garment Evaluation

The first-step investigation into the use of haptic-augmented garments for emotion regulation scenarios were conducted and presented in Chapter 6. There, the use of the haptic garment was in a ‘neutral to low arousal context’ (i.e., through a mindful meditation study) and was chosen as a starting point to understand the potential of the idea (i.e., it would not make sense to target a high arousal scenario if all participants are averse to the technology in a neutral or low arousal scenario). However, as mentioned, the long-term goal of the project is to employ such technologies to potentially counteract situations of acute stress—expanding upon groundwork laid by the occupational therapy community, using compression as an intervention for populations with sensory processing difficulties. The following proposes a study to understand the effects when this technology is used in a ‘neutral to high arousal situation’, i.e., when users are acutely experiencing a stressful situation. As before, due to restrictions imposed by COVID-19, the study has been postponed indefinitely.

Goal: Explore the use of computer-mediated compression to influence/mediate one’s emotions

Specific Aim: Determine if computer-mediated compression has the ability to induce calming effects in healthy young adults when exposed to acute stress

Method Overview: Within-subjects study, repeated baseline study design, with 2 Trier Social Stress Test [227] stressor tasks (one with and one without garment actuation; i.e., participants will be wearing the garment for both tasks, they just won’t know which one will actuate), presentation order of garment actuation is randomized/ counterbalanced. The overview of the study design is presented in Figure H-9.

Preference Baseline Stressor Stressor Recovery Recovery Interview Debrief Selection Relaxation Task 1 Task 2

Figure H-9: Haptic-driven emotion regulation (neutral-high arousal, with stressor tasks) study design overview

Measures: (1) Physiological data ▪ Empatica E4 (heart rate, electrodermal activity, temperature, etc.) ▪ Lab-developed breath sensor (breath rate) (2) Self-report measures (Pre-stress, Post-stress, Post-Recovery) 210

▪ State-Trait Anxiety Inventory (STAI) [209] ▪ Visual Analog Scale (VAS)- level of stress, happiness, feeling scared, calmness, irritation [228] (3) Qualitative Interview ▪ Perceived performance of task with and without garment actuation ▪ Garment comfort

Participant Selection and Study Information: ▪ Male participants, aged 18-34, size M ▪ Self-identified mentally and physically healthy, no prior cardiovascular health issues ▪ Target 20 participants ▪ Estimated study time: 2.25 hours ▪ Decieve participants’ with regards to study purpose o Participants will be told we are examining emotional responses to speech-related tasks and their relationship with wearable technologies. o Participants will also be told their task performance will be evaluated, rated against other participants, and video and audio recorded. ▪ Participants are instructed to get sufficient night’s sleep and avoid caffeine and alcohol the morning prior to testing

Study Script:

(Participants will be told to wear a t-shirt with long pants—to ensure the clothing worn by participants are generally similar) (Participants will be instructed to get sufficient night’s sleep and avoid caffeine and alcohol the morning prior to testing)

Introduction and Consent Process Estimated time: 10 minutes

“Thank you for agreeing to take part in our study. My name is (name of moderator). During the rest of the session, I will be working from a script to ensure that my instructions to every participant in the study are the same.”

“We are here to learn about individual’s emotional responses to speech-related tasks and their relationship with wearable technologies.”

“Before we start, we have a pre-test screening form that we need you to fill out, to make sure that you are eligible for the study.” (Provide pre-test screening form on Qualtrics to participant)

“Now, we have a consent form that we need you to read carefully. We will walk you through the document by providing you a brief overview of the study. Please do not sign the last page until we tell you so.” (Hand consent form to participant)

211

“In this experiment we will be testing your emotional responses to speech-related tasks and wearable technologies. With that, we will have you perform several speech-related tasks while wearing some wearable technologies. First, we have a wrist device which will capture physiological data including heart rate and electrodermal activity. While performing the speech-related tasks, you will also be wearing an upper-body wearable tech garment with instrumented sensors and actuators. You may feel some additional stimulus from the garment such as compression and warmth.”

“To give you an overview of what to expect in this session. First, we will let you experience the wearable technology garment and you will be asked your preference of stimulus application. For instance, where on the body you would prefer the sensations and in what intensity. This will also serve as a chance to let you experience the sensations by the garment, so you are not surprised by it later.”

“After that, you will perform two speech-related task, which I will explain right before you’ll be performing them. During this task, I ask that you perform them to the best of your abilities because your performance will be evaluated and rated against other participants. To do that, you will actually be audio and video recorded. However, your name will not be attached to any of that data and we only employ secure university cloud storage methods. Once the data is analyzed and transcribed, those audio and video files will be immediately deleted.”

“We will end the study with some surveys and interview about your experience. Again, your interview responses will be recorded and once your responses are transcribed, the audio files will be immediately deleted. I will also be taking notes throughout the study.”

“This session is estimated to take approximately 2 hours. At any point of the test, should you need to stop, please inform us immediately. Should garment become too uncomfortable, please let us know immediately and we will fix the problem or terminate the test. The risks of the speech tasks are minimal, but some people may be uncomfortable speaking in public. Should you choose to discontinue the study after the experiment begins, your compensation and your relationship with the university will not be affected.”

“To assess your understanding of what you just heard and read, in your own words, please briefly explain what you think is going to happen in the study.” (Check for understanding)

“Do you have any questions before we begin?” (Answer any questions that the participant may have)

“Thank you, you can now sign both of the consent forms. One of the copies is for you to keep.” (Collect consent form and leave one to participant)

Preference Selection Estimated time: 12 minutes ▪ After the consent process, participants are presented with the compression garment with multiple compression parameter options to determine each user’s preference. The possible parameter combinations with varying stimuli location options and intensity levels are presented below:

212

Parameter Options Low Intensity High Intensity Torso Shoulders Torso + Shoulders

▪ Each compression parameter option provided will be presented for 1 minute at constant compression level ▪ Each subject’s preference will be used as the compression garment stimulation parameter during the stressor task

“We are first going to start with fitting you with the wearable technology garment that will provide you with some stimuli, and we’re going to find the stimuli combination that you prefer the most.” (Help participant put breath sensor and garment on. The fit should be loose enough that it is similar to a regular t-shirt. Watch out for tight areas and fix them accordingly.)

“Are there any areas which you feel the garment is tight? Is the fit and tightness somewhat approximating a regular t-shirt? Are there areas that are uncomfortable?” (Fix garment accordingly and record responses/changes.)

“We will be selectively applying pressure on different regions of your body (the torso and the shoulder), and at different intensities (low and high). You will try each combination for about 1 minute and you should let us know what you preferred and didn’t like about each stimulus combination. At the end, you’ll provide us with a preference selection on the stimulus applied that you’ll be using for the rest of the study. For example, maybe you liked shoulder compression but not the torso and you preferred high intensity the most. Or you liked both shoulder and torso compression but at low intensities.”

“We’ll begin this exercise now. Starting with low torso intensity.” (Start with low torso, see if they want to increase intensity. Then add shoulder low intensity and see if they preferred addition of shoulder. If they preferred, give option of increasing intensity. Observe if they liked shoulder and torso combination. Choose to keep both or either one. Used a forced choice method—i.e., they have to pick something)

“Great! I’ll keep that setting and we’ll use these parameters for the later part of the study.” (Make sure garment is fully de-activated)

Empatica (biosignals data collection) Estimated time: 2 minutes

“Now, please put this device on your left wrist.” (Hand the Empatica to the participant and help participant put it on)

“This device is called the Empatica, it is a wrist-worn device used for collecting physiological signals in an easy manner, without the use of sticky electrodes on various parts of your body. This device collects heart rate and electrodermal information, which will allow us to monitor if there are physiological changes throughout the study. None of the collected data will be associated with your name, so you do not have to worry about any privacy issues.”

213

Baseline (Survey + Relaxation Adaptation) Estimated time: 23 minutes ▪ Pre-stress questionnaire: STAI, VAS ▪ Participants are told to sit quietly for 20 minutes (provided with magazines to browse)

“Now, I’d like you to fill out this baseline survey.” (Provide baseline survey—STAI and VAS form on Qualtrics)

“Before we start the main portion of the study, I’d like you to just sit here and relax for a little. Since you’ve been moving around, your physiological data will reflect those movements and potential emotional changes. So by having you sit quietly for a period of time, we are able to collect some baseline physiological signals. This period will last 20 minutes long. We have some magazines here for you to browse and pass time. I will inform you when the 20 minutes is up. Just take some time to relax and wind-down from the day.”

“I will need to press this event button on the Empatica to mark the start of this relaxation period.” *****(MARK START ON EMPATICA) ***** (Wait for 20 minutes)

“Okay the 20 minutes is up. Please let me press another event button to mark the end of this relaxation period.” *****(MARK STOP ON EMPATICA) *****

Stressor Task 1- Job Interview+ Math Estimated time: 15 minutes (3 minutes’ instructions + 12 minutes’ task) ▪ Empatica event marker (after instructions) ▪ Trier Social Stress Test (TSST)- Job Interview [229], [230] ▪ Empatica event marker (after task completion) ▪ Post-Stress: STAI, VAS

“Now we’ll move to the main portions of the study. In this task, we are going to ask you to give a short speech. We will be observing the speech, so that we can judge the quality of what you say as well as how you say it. You will be presented with a scenario involving a real-life situation that could happen to anyone. You will then be asked to make up a realistic story around the situation. I will describe the scenario to you and then you will have 3 minutes to plan the story and 3 minutes to speak. You may write notes during the 3-minute prep, but you will not be able to use those notes during the speech.”

“You have recently applied for a job that you think is perfect for your skills and is exactly the kind of job that you were looking for. You have been invited to come in for an interview to meet with several managers from the company. After arriving for the interview, please introduce yourself and then give a statement explaining why you are the right applicant for the job. Make sure to list specific examples that make you the best fit for the position.”

“You have 3 minutes to prepare your speech and then 3 minutes to deliver it. Please speak for the entire 3 minutes. Do not stop speaking until I ask you to. Do you have any questions?”

“I will need to press this event button on the Empatica to mark the start of this period.”

214

*****(MARK START ON EMPATICA) *****

“Please begin preparing now.” <<< Activate garment if ‘ON’ condition >>> (Leave participants in room for 3 minutes)

“Okay the three minutes is up and I will need to mark an event on the Empatica.” *****(MARK STOP ON EMPATICA) *****

“We have some audience members here who will be part of the team assessing your speech delivery.” (Audience enters.)

“It is very important for our research that you try your very best and that you speak for the entire time and look at the audience. If you do not try your very best, we will not be able to get good measures.”

“I will need to press this event button on the Empatica to mark the start of this period.” *****(MARK START ON EMPATICA) *****

“Please speak for the entire 3 minutes and do not stop speaking until I ask you to. You may begin speaking now.”

** If subject finishes early, respond with “You still have some time, please continue”

If the subject says that they have nothing additional to say, respond with: 1. What strengths do you bring to this job and what do you see as weaknesses in your application? 2. Tell me about a time when you have faced a difficult situation in a previous job. Please cite specific examples. a. How did you handle the situation? b. What was the outcome? c. How did you feel in this situation? d. What did you learn from this situation? e. Would you do anything differently if faced with a similar situation in the future?

If they still finish early, ask them to discuss what the hardest thing about job interviews is.

“That will be all, thank you. We will need to mark the end of this event on the Empatica again.” *****(MARK STOP ON EMPATICA) *****

“Next we are going to have you work on a series of simple arithmetic calculations. The mental arithmetic task is a test of your ability to perform a simple addition in your head.”

“To begin the task I will give you a two-digit number to start with. Your job is to add the two digits of that number, and then to add this sum to the two-digit number. FOR EXAMPLE, given the number 11, the sum of those digits is 2, plus 11 equals 13. So, given the number 13, the sum of those digits is what?” (Wait for a response) (Answer: 4) 215

“OK, then add the sum to the original number to get?” (Answer: 17)

“That is correct. So, your response would be like this: 13, 17, 25, 32 and so forth. You are to perform the addition as quickly and accurately as possible.”

“I will not interrupt you unless you make a mistake or until the 3 minutes are up. If you make a mistake, I will tell you to try again starting with your last correct response. I will be keeping a score and expect you to be able to accurately calculate each response in a certain number of seconds.”

“Please continue with the task for the entire 3 minutes. Do not stop until I ask you to. Do you have any questions?” (Answer any questions)

“Okay please let me mark this event on the Empatica.” *****(MARK START ON EMPATICA) *****

“Please begin now with the number 21” (Continue for 3-minutes)

“Okay the 3-minutes is up, thank you. We will need to mark the end of this event on the Empatica again.” *****(MARK STOP ON EMPATICA) ***** (Make sure garment is fully de-activated)

(Audience leaves)

Recovery Estimated time: 20 minutes* (*Literature found no significant difference between recovery of 15 minutes and 30 minutes, so estimated time spent of 5 minutes for filling up survey and 15 minutes of recovery) ▪ Empatica event marker ▪ Participants are told to sit quietly for 15 minutes (provided with magazines to browse) ▪ Post-Recovery: STAI, VAS ▪ Empatica event marker

“First I’d like you to fill out this survey.” (Hand Qualtrics survey)

“Now we’re at the recovery period, where I’d like you to just sit here and relax for a little again. Since you’ve been talking, this will help stabilize your physiological signals again. This period will last 20 minutes long. We have some magazines here for you to browse and pass time. I will inform you when the 20 minutes is up. Just take some time to relax.”

“I will need to press this event button on the Empatica to mark the start of this relaxation period.” *****(MARK START ON EMPATICA) ***** (Wait for 20 minutes) 216

“Okay the 20 minutes is up. Please let me press another event button to mark the end of this relaxation period.” *****(MARK STOP ON EMPATICA) *****

Stressor Task 2- Shoplifting Defense+ Math Estimated time: 15 minutes (3 minutes’ instructions + 12 minutes’ task) ▪ Empatica event marker (after instructions) ▪ Trier Social Stress Test (TSST)- Shoplifting Defense [230], [231] ▪ Empatica event marker (after task completion) ▪ Post-Stress: STAI, VAS

“In this task again, we are going to ask you to give a short speech. We will be observing the speech, so that we can judge the quality of what you say as well as how you say it. You will be presented with a scenario involving a real-life situation that could happen to anyone. You will then be asked to make up a realistic story around the situation. I will describe the scenario to you and then you will have 3 minutes to plan the story and 3 minutes to speak.”

“Imagine that you are shopping in a department store. You have been examining several articles to buy, as well as carrying a few items around with you while moving all around through many departments. A man in a business suit approaches you and identifies himself as the store manager. There is a policeman with him. In front of several passersby and others who are now looking on in interest – the store manager states that you have been reported by a plainclothes security staff to have shoplifted. You are becoming a bit embarrassed by the scene he and the officer are creating.”

“You feel you have been treated unfairly and were caused undue embarrassment in the store. Your task is to defend yourself against these accusations.”

“You have 3 minutes to prepare your speech and then 3 minutes to deliver it. Please speak for the entire 3 minutes. Do not stop speaking until I ask you to. Do you have any questions?”

“I will need to press this event button on the Empatica to mark the start of this period.” *****(MARK START ON EMPATICA) *****

“Please begin preparing now.” <<< Activate garment if ‘ON’ condition >>> (Leave participants in room for 3 minutes)

“Okay the three minutes is up and I will need to mark an event on the Empatica” *****(MARK STOP ON EMPATICA) *****

“Again, we have some audience members here who will be part of the team assessing your speech delivery.” (Audience enters.)

“Again, it is very important for our research that you try your very best and that you speak for the entire time and look at the audience. If you do not try your very best, we will not be able to get good measures.”

217

“I will need to press this event button on the Empatica to mark the start of this period.” *****(MARK START ON EMPATICA) *****

“Please speak for the entire 3 minutes and do not stop speaking until I ask you to. You may begin speaking now.”

** If subject finishes early, respond with “You still have some time, please continue”

If the subject says that they have nothing additional to say, please use the following prompts:

1. Please express how you would feel in this situation. 2. How do you think the others in the room would react to your defense? 3. What do you think should happen to you in this situation if your defense is not accepted? 4. Would you be willing to show the accusers what you have been carrying around the store, what is in your purse, pockets, other shopping bags, etc.?

If they still finish early, ask them to discuss what the hardest thing about job situation is.

“That will be all, thank you. We will need to mark the end of this event on the Empatica again.” *****(MARK STOP ON EMPATICA) *****

“Next we are going to have you work on a series of simple arithmetic calculations. The mental arithmetic task is a test of your ability to perform a simple addition in your head.”

“So, your response would be like this: 13, 17, 25, 32 and so forth. You are to perform the addition as quickly and accurately as possible.”

“Please continue with the task for the entire 3 minutes. Do not stop until I ask you to. Do you have any questions?” (Answer any questions)

“Okay please let me mark this event on the Empatica.” *****(MARK START ON EMPATICA) *****

“Please begin now with the number 14” (Continue for 3-minutes)

“Okay the 3-minutes is up, thank you. We will need to mark the end of this event on the Empatica again.” 218

*****(MARK STOP ON EMPATICA) ***** (Make sure garment is fully de-activated)

(Audience leaves)

Recovery Estimated time: 20 minutes ▪ Empatica event marker ▪ Remove garment ▪ Participants are told to sit quietly for 15 minutes (provided with magazines to browse) ▪ Post-Recovery: STAI, VAS ▪ Empatica event marker

“Now I’d like you to fill out this survey.” (Hand Qualtrics survey)

“We’re at the recovery period again, where I’d like you to just sit here and relax for a little again. This will help stabilize your physiological signals again. This period will last 20 minutes long. We have some magazines here for you again to browse and pass time. I will inform you when the 20 minutes is up. Just take some time to relax.”

“I will need to press this event button on the Empatica to mark the start of this relaxation period.” *****(MARK START ON EMPATICA) ***** (Wait for 20 minutes)

“Okay the 20 minutes is up. Please let me press another event button to mark the end of this relaxation period.” *****(MARK STOP ON EMPATICA) *****

Qualitative Interview Estimated time: 15 minutes

“Now, we will ask you a few questions to collect qualitative data based on your experiences just now. Just give us your honest opinion and all responses, positive or negative, are greatly appreciated.”

1. How did you find the overall experience? 2. What did you think the garment was for? 3. What terms would you use to describe the emotional experience of each of the two tasks? 4. Comparing the two tasks, did you notice any differences in the garment between the two? (If they say no difference: For one of the task, one of the garment was providing you some compression/warmth stimulus. Which one did you think it was?) 5. How would you say the garment actuation contributed to your experience? 6. How did you think the garment affected your performance? 7. Did you notice difference in emotions with and without the garment? 219

8. What did you like and dislike about the garment? 9. Anything else you’d like to share about your experience?

Debriefing Estimated time: 3 minutes ▪ Participants are told that their performance was not recorded and that the tasks which they were presented were unreasonably difficult and does not reflect upon their aptitude ability. ▪ The true nature of the study is on evaluating how haptics impact emotions during a stressor task.

“Thank you for your responses.”

“I wanted to let you know that the speech performance was not actually recorded. The tasks that were presented were unreasonably difficult and does not reflect upon your aptitude or ability. The true nature of the study is on evaluating how haptics—the wearable technology that you had, impact emotions and physiology during a stressor task. Apologies that we could not tell you of the nature of the task at the start because we did not want you to be biased towards the true purpose of wearable tech.”

“Do you have any other questions for me about the study?” (Answer any lingering questions)

“That concludes the entire study. Thank you for taking the time to perform this test with us. Your responses are appreciated and will help the team greatly in improving the usability of the garments. We would like to request that you do not discuss the test procedure and findings with anyone else to ensure that all the data collected are accurate and unbiased. Before we end, do you have any final questions for us regarding the study?” (Give participants gift card)

220