2 Personalized Learning, Personalized Space Upali Nanda, Phd, Assoc

PROJECT TITLE Personalized learning, personalized space

CONTACT Upali Nanda, PhD, Assoc. AIA, EDAC, ACHE Associate Principal and Senior Vice President Director of Research, HKS Architects Executive Director, Center for Advanced Design Research and Evaluation

712 Main Street Suite 1200 Houston TX 77002 +1 832 729 7652, [email protected]

RESEARCH TEAM Giyoung Park, PhD, AIA Angela Ramer Jon Bailey Tim Logan Jonathan Essary

INSTITUTION HKS Architects

SPONSORS Herman Miller USG

2 EXECUTIVE SUMMARY SUMMARY Objective This study aims to develop a sensory design lab (SDL) that measures human response (heart rate, anxiety, behavior) to interior design elements in real time. This study also intends to pilot-test the SDL in a local high school to measure the effects of furniture choice and arrangement and environmental conditions (sound, temperature, light) on heart rate, self-reported anxiety and achievement, and movement. Methodology The design team developed a manufacturer agnostic prototype for a portable, flexible and trackable lab – or the Sensory Design Lab (SDL). This prototype was installed in the DISD high School that supports Personalized Learning. A total of 30 students (up to three high school students at a time) participated in the experiment and performed a 30-minute independent work session in the 10’x10’x8’ SDL installed in the school. Students were instructed to set up the SDL using given four chairs, four stools, a table and a mobile whiteboard provided by Herman Miller. The chairs and the stools were located outside the SDL, and the table was folded at the beginning of each work session. Environmental sensors were used to capture sound, light, temperature and humidity. Thermal cameras were used to capture movement while maintaining privacy. Wristband fitness devices were used to capture heartrate and movement. A brief pre- and a post-experiment survey measured anxiety levels and self-reported achievement. Behaviors were coded manually. A personalization index was developed to assess the degrees of personalizing the space. Data was analyzed using SPSS for quantitative data and thematic content analysis for qualitative data. Findings Students in groups of three personalized the space more than smaller groups. Students who faced the entrance reported higher self-achievement than students facing the back wall. Chairs were preferred to stools, and stools were used as an additional surface for belongings. Participants often leaned back while seated. Among the environmental variables documented, sound levels were most salient. Overall anxiety levels reduced after spending the time in the SDL. Higher mean sound levels were correlated with less reduction in anxiety levels and lower achievement. Higher temperature and higher mean sound levels were associated with higher heart rate; yet, heart rate was not related to anxiety levels, self-reported achievement or behavioral measures. Greater fidgeting was associated with low minimum sound levels and with greater reduction in anxiety levels, but not with greater achievement. Implications Having visual prospect can contribute to cognitive performance; therefore, sufficient space allowing seating orientation change and a view can be beneficial. Larger groups’ greater degrees of personalization suggest that personalizable space may be more desired when in groups. There may be an

3 optimal range of sound levels in cognitive performance- not too loud which can cause anxiety, but also not too low that can cause students to feel restless. In this study, only furniture was changed as an independent variable. Sound level changed mainly due to conversations and ambient noise. But lighting and temperature stayed relatively constant. The success of the pilot shows the potential of systematically controlling light and temperature, finishes and colors, views and openings, proportions and many other interior design variables in future studies to assess the impact of interior design on human response. Originality and Limitations The SDL is a flexible, portable and trackable spatial system that aids design process. Its structure consists of brand-agnostic wooden beams with in- house 3D printed brackets and clips. Real-time measurement of environmental and human response in a set up that can be nimbly moved to project sites is original. The experimental protocol using a mixed method approach is replicable and can be significant for the field. Findings around optimal sound, and critical mass of students for personalization are also strong concepts that warrant further investigation. The student was a pilot prototype- and because of the lack of brand agnostic materials had to remain a rough prototype. Higher fidelity prototypes with more industry partners are warranted. Due to the constraints of time and budget ceiling and flooring were not included in the prototype and the volume could not be sealed. This could be remedied in the next version. Finally, the experiment produced vast amounts of data in different modalities. More efficient data collection and management are remaining challenges. It is important to remember that the sample size for the experiment was small, and the experiment should be replicated to be considered generalizable.

IMPLICATION Participants reported they appreciated the chairs provided because they HIGHLIGHTS could lean back. Casters not only allow students to move when needed more easily but also potentially reduce noise eliminating furniture dragging sound. Ergonomic aspects in furniture selection should be considered by schools. When sound levels were higher, reduction in anxiety levels was smaller during the work session, and self-reported achievement decreased. At the same time when it was too quiet students displayed more restless behavior. Acoustic design should be carefully considered to be “optimal” avoiding too loud or too quiet. Students in a group of three showed higher personalization index scores than the ones in a group of two or by themselves- this suggests that personalization may lend itself to a group dynamic more than individual

4 task work. Interior designers should consider larger level of personalization in areas where students will work in groups. Higher anxiety levels at the beginning of the experiment were associated with lower scores in personalization index. This implies that students who are anxious tend to personalize less. Students who faced the entrance wall reported higher self-achievement implying that elements of “prospect” may in fact help students feel a sense of achievement. Students reached for their backpacks during the work session or placed their belongings on a secondary seating. Space to locate belongings can help students’ immediate learning environment more organized.

RESEARCH BIO Upali Nanda, PhD, Assoc.AIA, EDAC, ACHE, Director of Research, HKS Architects (Principal Investigator) Dr. Upali Nanda is Director of Research for HKS, responsible for spearheading and implementing research projects globally. She also serves as the Executive Director for the non-profit Center for Advanced Design Research and Education. She is a member of the Academy of Neuroscience for Architecture (ANFA) Advisory Council, the AIA Research Advisory for Design& Health, and the AAH research council. Her doctoral work on “Sensthetics” has been published as a book available on Amazon.com, and she has published extensively in peer-reviewed journals and main stream media. Her research has been awarded the European Healthcare Design Research Award and two EDRA CORE awards. In 2015 Dr. Nanda was recognized as one of the top 10 most influential people in Healthcare Design for research, by the Healthcare Design Magazine.

Giyoung Park, PhD, AIA, LEED AP BD+C, Fitwel Ambassador, Senior Design Researcher (lead researcher), HKS Architects Dr. Giyoung Park is an environmental psychologist and registered architect focused on built environment, social capital, communication technologies, and human well-being. She has conducted both qualitative and quantitative research methods encompassing focus groups, interviews, online ethnography, virtual environment, experiments, structured and unstructured field observations, and surveys. Prior to her graduate training in human behavior and design at Cornell, she practiced architecture focusing on large commercial and healthcare projects in Bay Area.

Angela Ramer, Design Anthropologist/ Research Analyst, HKS Architects Angela’s background in anthropology brings a unique, humanistic and holistic research approach HKS design projects that highlight understanding the realities of human experience in built environments. Her consistent focus is on an engaged and empathetic understanding of user experience and expectations, helping to translate those insights into actionable interventions and accurate assessment of design. Angela graduated from the University of North Texas with a Master’s of Science in Applied Anthropology with a 5 concentration in business, technology and design. Her undergraduate education was completed at Elon University with a double major in Anthropology and International Studies.

Jon Bailey, Associate, Laboratory for Intensive Exploration (LINE), HKS Architects Jon has five years of experience in the architectural field, mostly in aspects of conceptual design. During his educational development, he was primarily focused on advanced digital modeling techniques, parametric and algorithmic modeling, and a focus on digital fabrication (primarily through the use of 6-axos robotic arms). The impetus for much of his design work revolves around the interaction and relationships between multiple systems, whether that is ecological, physiological, or building [systems].

Timothy Logan, Associate, Laboratory for Intensive Exploration (LINE), HKS Architects Tim has been working in the architectural field for 9 years, with much of his work focused on developing tools to help manage increasingly complex building designs and straddle the gaps between the multiple software tools used within the architectural practice. Gone are the days when pencils and straight edges are all the tools that architects need, and Tim has worked to ensure both geometry and data can be translated between different software. In addition to this software development work, he has also worked as computational designer for many project typologies including stadia, airports, hospitals, commercial buildings, working on anything from early design to fabrication.

Aaron Hollis, Architectural Designer, Education; HKS Architects Since joining the Education Group at HKS as an Intern Architect with a focus on applied research and technical applications, Aaron Hollis has assisted in over a dozen successful design projects, bridging the conventional gap between design and production processes. As a member of HKS’ ninth annual Design Fellowship, he worked closely with DISD in exploring the potential impact of personalized learning (PL, herein) upon the physical classroom environment and proposing design solutions to better facilitate this new program. He has continued to collaborate with DISD following the Fellowship’s conclusion- playing prominent roles in the design of three of DISD’s new PL campuses. Aaron graduated from the University of Texas at Austin with bachelor’s degrees in both architecture and architectural engineering in 2012, affording him a uniquely holistic background from which to approach both abstract and technical problems along with an array of rich multidisciplinary experiences particularly relevant to this new educational paradigm.

Jonathan Essary, Computational Designer/Lab Researcher, HKS Architects

6 Jonathan M. Essary is a computational designer/fabricator/lab researcher. His work includes investigations in spatial design, digital fabrication, digital analysis/design, architectural research, material research, research methodology, applied architectural theory, technology and its relation to human experience, and exploratory generative techniques to name a few reasons. Throughout graduate school he led point on a published research agenda regarding Ultra High Performance Fiber Reinforced Concrete in a Precast Façade Application, which was accepted and presented by Jonathan at the Façade Tectonics World Congress in October, 2016. His work is typically a pursuit of discovering the enigma of delightful space and a means to communicate that discovery. He holds both a BS in Architecture and an MArch degree from the University of Texas at Arlington.

ADVISORS ------Ashley Bryan, Director of Planning and Special Projects; Dallas Independent School District As a native of the Dallas-Fort Worth metropolitan area, Ashley grew up in and then taught at local public schools. Today, she continues to serve Dallas students as the Director of Planning and Special Projects in the Dallas Independent School District. In this role, Ashley primarily leads the district’s effort to implement PL district-wide and manages the Bill & Melinda Gates Foundation Next Generation Systems Initiative grant that helps to catalyze the work. She has also supported other cross-functional projects, such as summer school redesign and campus turnaround efforts. Prior to her current role, Ashley was an associate at Commit!, the Dallas County collective impact organization supporting cradle-to-career collaboration, where she focused on strategic initiatives across grades 4-12. She also taught 7th and 8th grade Spanish and spent several summers developing new teachers at Teach For America’s summer institute in Houston. Ashley graduated from The University of Texas at Austin with dual degrees in the Plan II and Business Honors programs and from the Harvard Graduate School of Education with a Master’s in Education in Education Policy and Management. Susan Whitmer. Lead Researcher Education; Herman Miller, Inc. As Herman Miller’s Lead Researcher for Education, Susan Whitmer combines her passion for creating great spaces to learn and work with research that impacts the future of the learning and teaching experiences. Susan holds a degree in design from Ringling College of Art and Design, a Master of Business Administration degree from Brenau University and a Master of Science in Accessibility and Inclusive Design from the University of Salford in the United Kingdom. Whitmer has written and co-authored numerous peer- reviewed white papers, articles, and book chapters on learning spaces, including Open House International, National Collegiate Inventors and Innovators Alliance, International Perspectives on Higher Education Research (Vol. 12), and SCUP’s Planning for Higher Education Journal.

7 Architecture and Design Advisors: Leonardo González Sangri, AIA, Vice President, Education; HKS Architects Leonardo González Sangri was born in Mexico City, Mexico. He graduated cum laude with a Master’s degree in Architecture from the Savannah College of Art and Design. Since moving to Dallas in 1998, he has worked in a broad range of building typologies, including educational facilities for both K-12 and higher education institutions, civic buildings, corporate campuses as well as being involved in urban planning for the City of Dallas. Grounded in his belief that educational spaces are of paramount importance in the successful future of our society, Leonardo has focused his career in the industry and has been designing community based projects such as the latest campus expansions of Uplift Education’s Charter Schools, assisting them in their mission of providing a high quality public education option in typically underserved neighborhoods. Leonardo’s efforts are aimed at advancing the building typology and creating spaces that enable new pedagogies and support progressive educational opportunities.

Transform Grant Final Deliverables: Detailed Research/Design Implications (May 24, 2017)

EXPERIMENTAL STUDY TITLE Personalized learning, personalized space: How design can enhance learning and overall wellbeing, to support the one-size-fits-one learning model

RATIONALE OF Personalized learning (PL) STUDY The competency-based or personalized learning (PL, herein) paradigm prepares students for college and career by implementing a more independent learning system tailored to individual student’s pace, educational needs, interests, and talents (U.S. Department of Education, n.d.). Greater flexibility in pedagogy and resources—including pace and time, location, learning materials and activities—can better facilitate PL paradigm. The model is implemented through three main strategies of 1) personalization (for the learner to drive his/her own learning in a competency based model), 2) differentiation (for the teacher to provide instruction to a group of learners and use assessment for learning), and 3) individualization (providing instruction to the individual learner based on individual needs). With the PL paradigm, a teacher and a pupil become partners for greater learning achievement and co-creation of curriculum. The effectiveness of this paradigm may be subject to the learning environment; yet, the relationship between the built environment and the outcomes of PL paradigm is unknown.

Environment quality and cognitive functioning Rapoport (1982) to an important connection between function and symbol in the physical environment and stated: “environments are more than just inhibiting, facilitating, or even catalytic; they not only remind, they also predict and describe” (p. 77). The environment “thus communicates, through a whole set of cues, the most appropriate choices to be made: the cues are meant to elicit appropriate emotions, interpretations, behaviors, and transactions by setting up the appropriate situations and contexts” (pp. 80-81). There is a compelling body of evidence showing that environmental design can influence students’ learning outcomes (Weinstein, 1981). In general, classroom acoustics are of concern; and both intensity and intermittence of sound can be influential (see Park & Evans, 2016 for review). If a substantial external sound source (e.g. aircraft) is not present, human-generated sound (e.g. chatter, dragging chair) is the primary noise source in the learning environment that can annoy both students and 1

teachers (Enmarker & Boman, 2004). Higher sound intensity and longer reverberation have been shown to significantly interfere with cognitive performance and social relationships between teachers and students and among students (Klatte, Hellbrück, Seidel, & Leistner, 2010). Chronic exposure to road traffic noise, on the other hand, can impair young children’s language acquisition (Cohen, Glass, & Singer, 1973) and elementary school children’s reading and math (Shield & Dockrell, 2008). In addition to cognitive performance impairment, chronic exposure to noise in the learning environment can lead to learned helplessness among nursery children in an empirical study where no significant traffic noise is present (Maxwell & Evans, 2000). Note that learned helplessness—belief that efforts will make no difference in outcomes—can decrease students’ motivation in learning and impair performance. In the same study, the children’s pre-reading and language skills as well as learned helplessness improved after installing acoustic panels in classrooms. Additionally, working in an environment that interferes with cognitive performance may lead to anxiety about performance. Anxiety can impair performance effectiveness (Eysenck, Derakshan, Santos, & Calvo, 2007). A short exposure to natural environment can help young adults restore from cognitive fatigue (Hartig, Mang, & Evans, 1991), children with ADHD concentrate better (van den Berg & van den Berg, 2011) and improve high school students’ mood and health (Han, 2008). Natural light and window views can support learning (Tanner, 2008) whereas highly decorative walls can distract students leading to more off tasking and impair learning outcomes (Fisher, Godwin, & Seltman, 2014, p. 1362). Studies have found that daylight, moderate temperature and humidity, and adequate air quality can improve learning outcomes (Barrett, Davies, Zhang, & Barrett, 2015; Schneider, 2002). Additionally, the attractiveness of a room influences positive affect and the energy level of those working in the room (Maslow & Mintz, 1956) and students’ learning process (Barrett et al., 2015). Low lighting, soft music, and comfortable seats encourage people to spend more time in a restaurant or bar (Sommer, 2007). Prospect- refuge theory suggests greater visual prospect and less visual exposure is preferred (Appleton, 1975). Moreover, personalized lockers or coat hooks, distinct characteristics of a room, ergonomic furniture, sufficient space and storage in the room, and designated breakout areas can aid students’ learning (Barrett et al., 2015). In a PL model, the role of the physical environment can be critical. Strategies that have been recommended and implemented to achieve this include (1) re-configurable, multipurpose rooms that can facilitate a range of learning modalities (e.g., project-based learning, teacher-led small-group instruction, inquiry-based instruction, blended learning), (2) collaborative student and staff spaces, (3) movable, multipurpose furniture, and (4) infrastructure to support anytime, anywhere learning.

The creation of activity or learning zones, studio-model classrooms, large communal learning spaces, and connectivity inside and outside spaces are other strategies that have been proposed (Rand Corporation, 2014; Steiner, Hamilton, Peet, & Pane, 20115). Due to the relatively recent focus on PL, few studies have investigated the effectiveness of these design strategies on the intended outcomes listed above. To encourage and enable school systems to invest in the substantial capital investment of renovating traditional spaces so that they support PL, such evidence is vital.

Personalized learning (PL) in Dallas ISD Dallas Independent School District (ISD) Context: Dallas, Texas has been recognized as the nation's second fastest growing economic region. As the 14th largest urban school district in the nation, Dallas ISD is in a unique and privileged position to plan for and implement this forward-thinking, PL-based model of education and create new spaces to facilitate 21st century learning. In the fall of 2013, Dallas ISD was among 20 districts across the nation to receive a grant by the Bill and Melinda Gates Foundation to develop a strategic vision for the development and implementation of PL models in local public schools. Following a successful central program-planning phase, Dallas ISD was awarded an additional grant to distribute directly to selected school teams committed to transforming their instructional model to support PL. After the development of PL program and training, five Dallas-area schools selected from the inaugural PL cohort launched their one-size-fits-one approaches to serving students in fall 2015, at three elementary schools (Cabell, Rogers, and Zaragoza Elementary Schools), one middle school (Marsh Preparatory Academy) and one high school (Innovation, Design and Entrepreneurship Academy at James W. Fannin, hereafter referenced as IDEA Fannin). These five schools have 1,584 students benefited from PL who represent DISD’s initial goal of about 1% of Dallas ISD students (current district enrollment is 158,787 students). By 2020 Dallas ISD aims to have 10% of all students in PL seats. Dallas ISD is also committed to holistically providing facilities and faculty to support this paradigm shift, which is encapsulated in the following objectives: 1) prepare students for a workforce that demands fluency in 21st century skills; 2) co-create education experiences alongside parents, the community, and students themselves within a broadly conceived education ecosystem, and finally 3) move beyond the limitations of the current system (DISD Strategic Plan, 2014). Prior to deploying the program, Dallas ISD leadership recognized that the existing physical conditions might not facilitate their new pedagogies in the most efficient manner. The Dallas ISD raised funds for design and construction in an “interim bridge plan,” approved by the Board of Trustees in March 2015 to address the most pressing facility needs 3

district-wide. Among the campuses selected for bridge plan projects, the PL sites were identified for renovations to support their evolving instructional paradigm. To kick off the project, a design charrette was conducted with an architecture firm, HKS Architects, and high school students, and key guiding principles were formulated. Renovations were planned for four of the five schools to create flexible learning environments to support a wide range of activities, from small group- teacher led instruction, to large collaborative project-based learning. Smaller break out rooms for one-on-one instruction, bars and nooks to plug-in for blended learning or independent study and a distributed library concept to provide ease of access to research materials were all designed in support of the PL model. Special care was placed to increase transparency into the learning areas and access to natural light. Spaces were crafted to support a variety of character traits, to ensure comfortable spaces cater to introverts and extroverts alike and allow everyone the opportunity to succeed along their PL path. Among the five PL schools, IDEA High School was selected for this study. The design and construction of the renovation was complete in 2016. Since the campuses began implementing the PL program this past August 2015, teachers have found various workarounds and have engaged in repurposing their existing environments, thus allowing a great opportunity for an ethnographic research on why space matters, and how (and why) teachers feel the need to change space to enable PL. It also gives us the opportunity to push the envelope on investigating how interior design can improve learning by establishing the first ever, sensory design lab in a grade-school setting.

Needs of real-time measurement system for a next level investigation We know, for example, that chronic exposure to higher intensity and intermittent noise can interfere cognitive performance and result in stress. However, sound measures many of such studies used were rough measures including Leq(A) contour maps (Park & Evans, 2016). These measures allow studies to use data from larger samples at the cost of accuracy. At the other end of the spectrum, studies focus on variables of interest in a controlled environment. This approach virtually removes other influential variables and examines one variable at a time. However, in real settings, contextual variables are always present and moderate the relationships between inhabitants and environmental variables of interest. Our approach is a hybrid of the two approaches. Real-time measurement of human-environment relationships in a semi-controlled real setting opens a next-level investigation in research. This approach development is briefly described in the method section, and its detailed history is included in the appendix A.

Significance of Proposed Research This project provided the opportunity to take a comprehensive, mixed- method approach to investigating the role of interior design in facilitating a PL program, and balancing qualitative and quantitative approaches. The project was conducted in partnership with the school system. The seed grant proposal for the installation of a sensory design lab (SDL) in a real educational facility provided students a new learning environment in their daily setting. The environmental and behavioral data gathered in the SDL would be closer to what could result from their real learning environment compared to the ones from traditional lab studies. Furthermore, this study provides a new method of baseline condition assessment for future learning environment design. Findings of the present study about how interior design elements can impact learning outcomes can help not only Dallas ISD but also the US education system with spatial prototypes and design implications.

PURPOSE OF The SDL study aims to develop and pilot the SDL that measures human STUDY response to interior design elements in real time. The pilot test explored the relationships among 1) furniture selection and arrangement, 2) ambient environments (sound, illumination, temperature and humidity), 3) human behavior (e.g. distractions, interface changes), 4) learning achievement, 5) anxiety levels, and 6) heart rate.

METHODOLOGY Sensory Design Lab development (SDL) The team developed a pop-up sensory design lab. The lab is a room-in- room system that is to portable, trackable, and flexible. The size of the lab, 10’W x 10’ W x 8’H, was determined by the room where it would be installed in the school. The lab is equipped with five custom-made environmental sensors measuring illumination, sound levels, temperature and relative humidity every few seconds; two thermal cameras to document inhabitants’ behavior while protecting their privacy; and fitness wristbands recording each participant’ heart rate. Herman Miller provided furniture pieces suitable for learning environments. See Appendix A for a detailed report on how this lab was developed as well as future implications.

Data collection tools and instruments Environmental data: Five environmental sensors were installed in the SDL to document sound levels, illumination levels, temperature and humidity every five seconds (Figure 1a). Wristband device: Movisen was considered for heart rate measure when the study was proposed, but this would require chest strap. The team

decided to use a wristband fitness device instead for its less invasive interface. The device measured the number of steps. Thermal cameras: Two thermal cameras were installed in the SDL. The researchers developed behavior coding criteria, pilot-tested a few times, and a trained researcher coded all thermal video recordings. Coded behaviors include 1) sit and stand, 2) laptop, phone, and printed material use, 3) social interaction, 4) whiteboard use, 5) movement of furniture pieces, and 6) signs of distraction such as fidgeting and looking out. In most cases, we could distinguish whether a student used a mobile phone for school work or off-tasking. Therefore, we coded this as well. Figure 2. Sample datasheet of aggregated environmental, behavioral and heart rate measures per minute.

Occupancy sensor data: The occupancy sensor data from each furniture piece measured how long the furniture was used in a 10-minute interval. Pre- and post-experiment surveys: A pre-experiment survey asked anxiety levels, goals to achieve during the experiment, and potential interrupters to the goal achievement. A post-experiment survey measured anxiety levels again using the same instrument, how much was achieved, and actual interrupters to their work. The researchers verbally asked the participants i) why they set up the space in a certain way, ii) what other environmental qualities and equipment would help their learning, and iii) what else they wished to have in their learning environment. See Appendix B for a copy of pre- and post-experiment surveys. STAI instrument: Anxiety levels were measured using a six-item, four-point Likert-scale STAI instrument (Marteau & Bekker, 1992, =.82). The scores would range from 20 to 80. Self-reported achievement: A post-experiment survey included a α five-point Likert scale question, “To what extent did you achieve the goals you had set for yourself?”

Experiment setting The SDL was installed in a small conference room in the school next to the flex space. Four task chairs, four hexagon stools, a 36”-square foldable table, and a mobile whiteboard were provided by Herman Miller. All furniture pieces except the stools had casters. The chairs and the stools were located just outside the SDL and in the conference room so that participants would need to locate them. The folded table and the white board were located by the wall in the SDL due to limited space in the conference room. An occupancy sensor was attached to each furniture piece. Four of the five environmental sensors were attached to the SDL’s interior walls, and the other one was installed on the table top. The baseline layout was held consistent during the experiment. Figure 1. The experiment setting: a) base-line furniture and sensor layout and b) a picture of baseline condition. Thermal cameras were installed at upper right and lower left corner of the diagram a).

Experiment procedure Upon the grant award, RRB approval from Dallas ISD was sought. Parental consent was collected at the beginning of the academic year after RRB approval. Only consented students were invited during their flexible learning lab period. Up to three students were allowed per session. After filling out the pre-experiment survey, participants were given a fitness wristband and asked to set up the lab space with furniture pieces provided any way they wished. Then, they worked on tasks they brought

for the next thirty minutes. After the work session, the participants answered the post-experiment survey. The second page of the post- experiment survey in Appendix B were asked verbally by the researchers. A picture of furniture layout was taken when participants were filling out the post-occupancy survey.

Analysis methods The present study developed two indices to assess students’ behavior including personalizing the space. Interface index: Students often changed interface from one to another. This index indicates how many times such interface change occurred, aggregating the frequencies of laptop, phone, tablet, hard copies, whiteboard, and in-person conversations during the work session to explore if interface change was related to other outcomes. Personalization index: Furniture set up and layout were captured by post- experiment photographs. The configuration was coded with several items such as orientation of seat and visibility to outside using pictures before and after the experiment. Additionally, a personalization index was created using the sum of five binary items (0–5): 1) whether a table was used (1) or not (0); 2) whether the table was moved (1, moved away from its baseline location) or not (0, not used or moved only to unfold the table) 3) whether an additional seating was taken in to the lab (1) or not (0); 4) whether a whiteboard was used (1) or not (0); 5) whether the white board was moved (1) or not (0). While the environmental sensors generated snapshot data (e.g. temperature of a given m oment), wristband sensor data were aggregated data per minute. Therefore, in this pilot test, behavior and environmental data were also aggregated and combined with the wristband data. Whereas temperature and humidity data were equivalent among the sensors, illumination and sound levels varied significantly. It is possible that the illumination levels varied substantially by location and by other factors such as penetrating daylight. In this pilot test, the two illumination- level measures, one from the table and the other from an unblocked wall, were used as most student used the table as work surface. The sound levels, on the other hand, were not expected to vary much by location. Even though they had been calibrated twice prior to data collection, two of the five sensors’ sound levels were noticeably different. Therefore, sound levels from only one sensor were used in the analysis. Five sound measures were generated when aggregating the environmental data: maximum, minimum, maximum-minimum, mean sound levels and standard deviation. Similarly, illuminations levels 8

generated: mean illumination levels on the work surface, mean illumination levels from the sensors on the walls, and standard deviation of work surface illumination levels. Those who did not use the table, the work surface lighting measures were removed. Illumination levels were insignificant in the analyses thus will not be discussed in the results; and humidity has not been analyzed yet. The occupancy sensory data did not add more information to the behavior coding, and the number of steps measured by the wristband device had missing data. Therefore, neither occupancy nor the number of steps was used in this study. This study generated one qualitative dataset and two aggregated quantitative datasets explained below. Histograms and correlation tests were generated to identify patterns among variables. Figure 3 visualizes multiple datasets during an experiment. Further statistical analyses were conducted as described here. 1) Pre- and post-experiment survey: The quantitative portion of the surveys was transferred to an excel file and combined with other datasets below. The qualitative portion was analyzed using thematic analysis and content analysis methods (Vaismoradi, Turunen, & Bondas, 2013). 2) Per-person dataset: Behavioral data were aggregated per person and combined with pre- and post-experiment surveys and the personalization index scores. Analysis of variance (ANOVA) tests were performed to see differences among groups—such as groups of 1-, 2-, and 3-persons. When appropriate, generalized estimating equations (GEE) test was used. GEE is a type of regression analysis method that can consider some participants are in groups. 3) Per-minute dataset: Behavioral, environmental (sound, illumination, temperature), and heart rate data were aggregated per minute, per person (Figure 2). Each person has multiple measures, and individual measures are nested in group levels. GEE tests were used as it is also suitable for analyzing repeated measures. Figure 3. Environmental, behavioral and heart rate data of a group session. Two participants’ behavior and heart rates are overlaid. Conversation, hard copy, laptop, phone and whiteboard use are marked with dots when the usage started. The sum of the dots is the person’s interface index score. Person 1’s data are the upper rows of the behavior and BPM; and person 2’s data are the lower rows.

Gender and group size were controlled for in the quantitative analyses when appropriate as they might influence the outcomes. Variables were removed in the final model if not statistically significant (<.05).

KEY FINDINGS • Majority of students (28 out of 30) chose a chair over a stool for task performance; and chairs were more likely moved (20 of 28, M=3.67, SD=5.39) after initial set up, probably due to ease of movability via casters. About half of the sessions (8 out of 17) used a white board.

• The groups of three (M=3.22, SD=.67) showed higher personalization index scores than the ones of one (M=2.14, SD=1.22) or two students (M=2.00, SD=1.11) [F(2, 27)=4.175, p=.026]. However, personalization index scores were not related to self-reported achievement. • Students who faced the entrance wall reported higher self- achievement (M=4.25, SD=.71) than those towards the back wall (M=3.00, SD=1.00, N=11) (p=.042, 95%CI=.053, 2.447). Greater visibility from the seat may be positive, which is aligned with prospect-refuge theory (Appleton, 1996). • Participants’ anxiety levels in pre-experiment (M=35.44, SD=9.84) were higher than in post-experiment (M=28.33, SD=7.36) [t(29)=4.307, p=.000]. The reduction in boys’ anxiety levels during the experiment (M=-12.73, SD=10.63) was greater than girls’ (M=-3.86, SD=6.21) [F(1, 28)=8.409, p=.007]. • Personalization index scores and anxiety level change were not associated. However, anxiety levels at the beginning of the experiment was associated with personalization index scores implying students with higher anxiety levels personalized the space to a lesser degree ( =-.066, SE=.020, 95% CI=-.106, -.026, p=.001). • Higher sound levels were linked to less reduction in anxiety controlling for groupβ size ( =2.010, SE=.864, 95% CI=.317, 3.703, p=.020). With one decibel decrease in mean sound levels, reduction in anxiety levels decreased by 2β points. The relationship between anxiety levels and noise in learning environment has not been identified in previous studies; yet, this finding appears to be aligned with the relationship between sound levels and self-reported achievement in this study. • Higher mean sound levels were associated with lower higher self- reported achievement ( =-.091, SE=.029, 95% CI=-.147, -.035, p=.001), which is consistent with previous studies’ findings about sound levels and learning outcomes β(Park & Evans, 2016). With a 10 dB increase in mean sound levels, self-reported achievement score decreased by .9 point in a 1-to-5-point scale (M=3.63, SD=1.03, N=30). Noise can interfere with both teachers and students—both teaching/learning and their social relationships (Enmarker & Boman, 2004; Klatte, Hellbrück, Seidel, & Leistner, 2010). Note that chronic exposure to noise in learning environment may result in long-term effects including learned helplessness (Maxwell & Evans, 2000). • Students’ mean heart rate ranged from 60 to 109 per minute (M=90.46, SD=12.98, N=25). Higher temperature was linked to greater heart rate ( =5.67, SE=2.09, 95% CI=1.57, 9.77, p=.007). And, higher maximum sound levels were associated with greater heart rate ( =.312, SE=.122, 95%β CI=.073, .551, p=.011). However, heart rate was not related to anxiety levels or change in anxiety levels. Heart rate may not beβ a reliable indicator of small to moderate anxiety level changes. Blood

pressure or facial recognition technology for mood measure may be considered for future studies. • Fidgeting was linked to low minimum sound levels (baseline sound levels) ( =-.122, SE=.394, 95% CI=-.199, -.045, p=.002) implying there may be an optimum sound level range and/or types of sound supportingβ cognitive performance. Fidgeting was not associated with mean sound levels. Greater fidgeting and greater phone offtasking were also related to greater reduction in anxiety. • The interface change index showed students frequently changed interface (M=15.03, SD=8.71, N=30) yet was not associated with other measures.

Design implementations • Participants reported they appreciated the chairs provided because they could lean back (Barrett et al., 2015). • Casters not only allow students to move when needed more easily but also potentially reduce noise eliminating furniture dragging sound. • Some students took in additional seating to put their belongings on, and several reached for their backpacks during the work session. Space to locate belongings nearby can help students’ learning (Barrett et al., 2015). • Acoustic design should be carefully considered to reduce negative effects of high sound levels. Unless substantial external noise source (e.g. aircraft) is present, human-generated noise in the learning environment is the primary noise source (Enmarker & Boman, 2004). • Too quiet environment can result in fidgeting. A white noise generator may be considered to optimize background sound levels while masking interfering noise.

REFERENCES Appleton, J. (1975). The experience of landscape. Chichester: Wiley. Barrett, P., Davies, F., Zhang, Y., & Barrett, L. (2015). The impact of classroom design on pupils' learning: Final results of a holistic, multi-level analysis. Building and Environment, 89, 118- 133. doi:10.1016/j.buildenv.2015.02.013 Cohen, S., Glass, D. C., & Singer, J. E. (1973). Apartment noise, auditory discrimination, and reading ability in children. Journal of Experimental Psychology, 9(5), 407-422. doi:10.1016/S0022-1031(73)80005-8 Dallas ISD. 2014. Personalized learning in Dallas ISD strategic plan. Dallas, TX. Enmarker, I., & Boman, E. (2004). Noise annoyance responses of middle school pupils and teachers. Journal of Environmental Psychology, 24(4), 527-536. doi:10.1016/j.jenvp.2004.09.005 12

Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive performance: attentional control theory. Emotion, 7(2), 336-353. doi:10.1037/1528- 3542.7.2.336 Fisher, A. V., Godwin, K. E., & Seltman, H. (2014). Visual environment, attention allocation, and learning in young children. Psychological Science, 25(7), 1362-1370. doi:10.1177/0956797614533801 Han, K.-T. (2008). Influence of limitedly visible leafy indoor plants on the psychology, behavior, and health of students at a junior high school in Taiwan. Environment and Behavior, 41(5), 658-692. doi:10.1177/0013916508314476 Hartig, T., Mang, M., & Evans, G. W. (1991). Restorative effects of natural environment experiences. Environment and Behavior, 23. doi:10.1177/0013916591231001 Klatte, M., Hellbrück, J., Seidel, J., & Leistner, P. (2010). Effects of classroom acoustics on performance and well-being in elementary school children: A field study. Environment and Behavior, 42(5), 659-692. Marteau, T. M., & Bekker, H. (1992). The development of a six-item short-form of the state scale of the Spielberger State Trait Anxiety Inventory (STAI). British Journal of Clinical Psychology, 31(3), 301-306. doi:10.1111/j.2044-8260.1992.tb00997.x Maslow, A. H., & Mintz, N. L. (1956). Effects of Esthetic Surroundings: I. Initial Effects of Three Esthetic Conditions Upon Perceiving “Energy” and “Well-Being” in Faces. The Journal of Psychology, 41(2), 247-254. doi:10.1080/00223980.1956.9713000 Maxwell, L. E., & Evans, G. W. (2000). The effects of noise on pre-school children's pre-reading skills. Journal of Environmental Psychology, 20(1), 91-97. doi:10.1006/jevp.1999.0144 Park, G., & Evans, G. W. (2016). Environmental stressors, urban design and planning: implications for human behaviour and health. Journal of Urban Design, 21(4), 453-470. doi:10.1080/13574809.2016.1194189 Rand Corporation. (2014). Early progress: Interim research on personalized learning_2014. Retrieved from http://collegeready.gatesfoundation.org/wp- content/uploads/2015/06/Early-Progress-on-Personalized-Learning-Full-Report.pdf Rapoport, A. (1982). The meaning of the built environment: A nonverbal communication approach: University of Arizona Press. Schneider, M. (2002). Do school facilities affect academic outcomes? Retrieved from Washington DC: http://www.ncef.org/pubs/outcomes.pdf Shield, B. M., & Dockrell, J. E. (2008). The effects of environmental and classroom noise on the academic attainments of primary school children. The Journal of the Acoustical Society of America, 123(1), 133-144. Sommer, R. (2007). Personal space; Updated, the behavioral basis of design. Bristol: Bosko Books. Steiner, E. D., Hamilton, L. S., Peet, E. D., & Pane, J. F. (20115). Continued Progress: Promising Evidence on Personalized Learning. Retrieved from http://www.rand.org/pubs/research_reports/RR1365z2.html Tanner, C. K. (2008). Explaining Relationships Among Student Outcomes and the School's Physical Environment. Journal of Advanced Academics, 19(3), 444-471. doi:10.4219/jaa- 2008-812

U.S. Department of Education. (n.d.). Competency-based learning or personalized learning. Retrieved from http://www.ed.gov/oii-news/competency-based-learning-or- personalized-learning Vaismoradi, M., Turunen, H., & Bondas, T. (2013). Content analysis and thematic analysis: Implications for conducting a qualitative descriptive study. Nursing & Health Sciences, 15(3), 398-405. doi:10.1111/nhs.12048 van den Berg, A. E., & van den Berg, C. G. (2011). A comparison of children with ADHD in a natural and built setting. Child Care, Health and Development, 37(3), 430-439. doi:10.1111/j.1365-2214.2010.01172.x Weinstein, C. S. (1981). Classroom design as an external condition for learning. Educational Technology, 21(8), 12-19.

Appendix A The Evolution of Sensory Design Lab

Structure Selection Initial constraints placed on the sensory lab included that is must be lightweight, flexible, modular, brand agnostic, and within our allowable budget. Our early investigations and search for products already existing on the market led us toward a few different room-in- room systems. These (room-in-room) systems are made from extruded aluminum profiles with the ability for wall panels to clip in. These systems range in price, but greatly exceeded our budget. We found only one vendor who had a room and were willing donate the room system in-kind. However due to volume constraints at our experiment location several modifications would need to be made. Furthermore, these systems are typically not designed structurally to support a ceiling systems, they typically tie in to an existing ceiling system. Additional challenges with the room-in-room systems were the requirement for licensed installers to take down, transport, and put back up the system. If a member of our team wanted to be able to do this, they would need certification for installation from the company. Though many room-in-room systems exist on the market, many (if not all) require a propriety wall to frame clipping mechanism, where wall panels are typically fitted with a bolt in the back of the panel and then joined to the frame. In the case of our system we need the flexibility to change out wall panels without modifying them (cutting or fastening permeant bolts in the back).

All of these factors led our team to explore the creation of our system using extruded aluminum profiles. While it is possible for our team to build the portable lab out of extruded aluminum shapes there were concerns of being able to accurately modify the shapes by cutting in our in-house model shop.

Due to cost and time implications our team decided to look at a modular wood frame system made from Grid Beam. This system met all of our criteria and was a system that was easily modified in-house. In our experiment, we also wanted the interior to be comprised of flush wall panels, similar to typical interior construction practice. Variables of a flush detail and having the ability to accept multiple wall panel thicknesses required the team to explore multiple variation of corner bracket details and wall panel clips.

Takeaways

1. While the wood framing systems met the criteria of being lightweight, modular and easily customized, issues of long-term durability and overall structural [in]ability to support heavier wall panels and a ceiling remains a limitation of the system. a. These limitations on the wood framing system are areas that could be addressed through an extruded aluminum framing system. 2. Another desire of the lab is to be able to adjust variables such as lighting levels and perhaps even thermal comfort in the room. Due to the limitations of the wood framing system of being able to support a ceiling system, there is a lack of ability to support lights and HVAC registers. While the goal of constrained volume is to make sure the volume remains a constant variable, if the desire was to fully enclose the volume with a ceiling, there would be an inability to properly light and ventilate the space within. Appendix A

Framing Vendor Cost Pros Cons Type Aluminum MiniTec $6,125 for frame • Vendor Design Assist • Potentially not possible to Extrusion Or and a sliding • Brand Agnostic put a ceiling, need to work Modular door. No wall or • Finished Look with engineers Framing, ceiling panels. • Modular construction • Need to design clips to fit FramingTech, • Proprietary locking wall panels flush on the Grainger mechanism interior. • No licensed installer • Modifications need to necessary happen in the factory. • Sliding acrylic door • Walls cannot support much available weight (open mesh grid, acrylic, lukabond) • Infill panels, or hand screw fastened Modular GridBeam 112' x $8/ft = • Flexible (2', 4', 6', and 8' or • A rougher or unfinished Framing $896 plus custom size) aesthetic (Wood, shipping • Lightweight • May not be able to support Aluminum) (aluminum) 1- • Modular construction much weight (as wood) 1/2"w beams, system • No sliding door option with $9/ft for 2"w • Within price range wood (reqs. Steel beams beams • No licensed installer from UniStrut) 112' x $3/ft = necessary, free consulting • No ceiling opportunities: no $336 plus available load bearing info available shipping for wood or aluminum (shipping costs • 8' limit for freight shipping higher when • Need to work on finish using 8+' attachment detail, elements for especially glass doors. 1/8" cleaner look.) for plexi recommended (4'x8' 10'x10'x8' max available) Metal UniStrut 2" beams (same • Modular lightweight system • Heavy (2 lbs / ft) Framing as hole spacing) • Cleaner edges than • Need to work on finish $5/ ft according gridbeam due to tubing attachment detail, to gridbeam system and joint detail especially glass doors website • Suppliers to GridBeam [metal] system • Local rep Plastic EverBlock $5,700, includes • Modular construction units • No ability to apply walls or Blocks shipping • Lightweight ceilings to EPS Foam N/A $600-$800 ea. • Lightweight • No fabrication equipment 4x4x8ft • Flexible to modify blocks (CNC or • Can be crafted to give Robotic Arm) texture, form, etc. • Expensive, including shipping + handling Cardboard N/A • Lightweight • Unfinished look • Flexible • Not a realistic version for materiality • Not a framing system, unless combined with something else or more of itself to create structural ability Appendix A

Room-In- Teknion $2500-$3,000 + • Solution requires the least • Not brand agnostic, need Room shipping/install work from HKS for DISD proprietary clipping • Vendor Design Assist mechanism • Not flexible past DISD project • Requires licensed installers • All modifications happen in factory • Height of current system is either too low, or too tall for DISD space. Room-In- DIRTT $150,000 • Universal clipping system • Out of our price range Room • Political disputes No N/A Cost of sensors • “Lab-in-a-bag”, does not • Does not meet the Framing require framing system. requirements of the ASID System proposal for a portable design lab. Custom Local • Could be fabricated fairly • Potentially a very heavy Steel Fabricator quickly system Frame • Likely would have no flexibility, unless more design time was put into the system Post and Steelcase • Freestanding • Potentially not brand Beam • Vendor Design Assist agnostic • Similar to Teknion, typical Room-In-Room systems • Potentially a heavy system, not very transportable • Potentially requires licensed installers Clean Terra $12,000, • Can have built-in HVAC and • Least flexible and Room Universal 10x10x7 lighting transportable of the System w/acrylic walls • Can have a suspended options. ceiling system • Requires 2'-0" of clearance • ISO 5 Hepa filtration system overhead • Can control lighting + air quality/composition

A sensory design lab 2.0 is considered as a higher fidelity prototyping structure with the ability to closely replicate typical interior finishes while still providing a lightweight, flexible, trackable, and portable system for interior spatial prototyping. Further development of the structural system would benefit greatly from an upgrade in materials to extruded aluminum members as a higher fidelity second iteration. Upgrading from wood to aluminum provides a performance increase in many areas including strength, flexibility, rigidity when built, cleaner interior finish, versatility, and integration of other technology such as touch screens, full height walls screens, green walls, immersive cave projection and sound systems, and kinetic installations, while maintaining portability with minimal weight increase. Concerns addressed in the built prototype such as quick interchangeability of panels and modular assembly of components which made the wooden beam members the system of choice provide an insight into the viability of an aluminum system previously thought to be challenging or unattainable. The lessons learned from the current prototype increase confidence in the viability of an aluminum system. After investigation into different systems both existing room-in-room systems and reconfigured Appendix A agnostic systems, we selected a custom extruded aluminum framing system (MiniTec) for our next generation of the SDL system. Among several extruded aluminum systems available MiniTec provides an affordable yet more customizable system. Some key options include multiple types of panel attachments, many options for structural connections, shop services for custom lengths, door options both hinged and sliding, pre-threaded end bore options, and wide members for increased bending strength providing the option for a ceiling with the appropriate design. The key differentiator within affordability is the options provided for customization both with fasteners, profile types, and the ability to add sections within a built structure due to the space provided in the slot detail of the profile types. This system is utilized in industry builds from engineering prototypes, to machine enclosures, to room enclosures and has a proven record of consistent finished results, structural dexterity, and customizability while maintaining a brand agnostic solution.

Sensor Selection Several ready-made sensors were evaluated, particularly the Bosch XDK (donated by USG) and the Texas Instruments SensorTag. Both had great battery life and possible Bluetooth/WIFI connectivity. The Bosch sensor was the most advanced, in that it had all of this plus local data storage capabilities, whereas the SensorTag would need a host to connect to that could log the sensor data.

BOSCH XDK Texas Instruments SensorTag Custom Arduino Based PROS • Local Storage for Data • Small Size • Local Storage for Data • Battery Life (up to 1 month) • Battery Life (up to 1 year) • Accurate time without WIFI • Many Built In Sensors • Many Built In Sensors • Expandability/ Variability • Wireless connectivity • Inexpensive • Inexpensive • Wireless connectivity • Wireless connectivity CONS • Price • No Local Storage • Battery Life (up to 2 days) • No Expandability/ • No Expandability/ • Expandability/Variability Variability Variability • Testing needed for sensor • No timestamp without WIFI • No Audio Sensor accuracy • No Timestamp without WIFI

In the end, we went with the custom Arduino based option primarily for the expandability/variability the platform allows. It was most fitting with the spirit of the project. Cost was another aspect, with the SensorTag being overwhelmingly cheaper option (approximately $30), but would require some other computer system (even a cheap [<$100] Appendix A

Linux machine like a Raspberry Pi or BeagleBone) for datalogging. The Arduino based approach gave us the flexibility to design the sensor however we saw fit.

We purchased two Flir thermal cameras to document students’ behavior during the experiment while protecting their privacy. Additionally, we selected a fitness wristband, Garmin Vivosmart HR+ because we could download raw data from its dashboard unlike other equivalent gadgets available on market. Movisen was initially considered and proposed in the proposal. However, it requires chest trap for heart rate. We decided to use a wristband for its less invasiveness. The device measures heart rate, number of steps, calories consumed, and floors climbed. Having raw data would allow us to sync the data with other datasets.

Takeaways

1. Sensors were programmed to take a data sample every 5 seconds, but that 5 second interval was dependent on when the sensor was turned on, so it became unlikely the data samples would occur at a specific timestamp. Having it log at specific second intervals would provide more consistent data across the board. 2. More time devoted to case design, and testing how the case design may affect sensor data acquisition. 3. Selection of sensors was largely based on cost, but ideally we should have purchased examples of many sensors and evaluated best fit accordingly. 4. More time for testing calibration before deployment would have been helpful. Because lab design and solving it’s problems were in the forefront for a large part of the development process, the sensors were not given the same time to mature. 5. A more expensive set of base components (Arduino, SD Card writer, Real-Time Chip [RTC]) were selected due to the desire to have WIFI/Bluetooth connectivity for possible expansion even though it wouldn’t be utilized in this first test run. Choosing a different set of components could have lowered the cost and upped the RTC accuracy while likely also providing better battery life, but the components would have been relegated to local data loggers only without the addition of another costly or space consuming wireless module (WIFI, Bluetooth, radio) and could have still necessitated the need for a separate host akin to the SensorTag (Raspberry Pi/Beaglebone type computer). 6. Behavior coding using the thermal camera recordings took a full week after coding criteria development. 7. Due to the small volume of the SLD and the room where SDL was installed, two thermal cameras were not able to document the entire SDL. An alternative we considered was using typical wide angle camera with a filter to obfuscate identities. We are also looking for other options to capture wider areas and to save time in behavior coding. 8. The device needed to be set as ‘active mode’ to increase the accuracy of movement and the frequency of data collection, which was overlooked during the experiment. The movement data were not accurate thus not used in the analyses.

Data Management All datasets were imported into Microsoft Excel. This pilot test generated nearly 30,000 rows of environmental data, about 500 rows of Garmin data, and about 1,000 rows of Appendix A behavioral data. Garmin device data were not snapshot data like the environmental and behavioral data. The behavioral data had when each behavior started and ended (e.g. laptop use) or when it occurred if the behavior would not last long (e.g. moving a chair). All the data, except the one from furniture occupancy sensors, were aggregated data per minute. The furniture occupancy sensors aggregated all data of each 10-minute segment We discussed data-syncing options and decided to aggregate environmental and behavioral data per minute to combine the data with the Garmin data. By doing so, the dataset loses granularity of constantly fluctuating sound and illumination levels. To compensate this, multiple sound and illumination measures were created such as mean, maximum, minimum and standard deviation values. Instead of using the dataset, we coded sit and stand behavior using the thermal cameras.