SOCIOTEMPORAL DISORDER IN GLOBALLY DISTRIBUTED DIGITAL HUMANITARIAN WORK

by

Wendy Norris

B.A., University of South Florida, 2010

A thesis submitted to the

Faculty of the Graduate School of the

University of Colorado in partial fulfillment

of the requirements for the degree of

Doctor of Philosophy

Department of Information Science

2020

Committee Members:

Dr. Stephen Voida, University of Colorado Boulder

Dr. Leysia Palen, University of Colorado Boulder

Dr. Amy Voida, University of Colorado Boulder

Dr. Brian Keegan, University of Colorado Boulder

Dr. Katrina Petersen, Trilateral Research ii

Norris, Wendy (Ph.D., Information Science)

SOCIOTEMPORAL DISORDER IN GLOBALLY

DISTRIBUTED DIGITAL HUMANITARIAN WORK

Thesis directed by Dr. Stephen Voida

“What time is it?” is a deceptively simple question. Sociotemporality, or the social experience of time, is often taken for granted until there is a coordination breakdown. This dissertation ex- plores an especially acute breakdown in time: the high-tempo, time- and safety-critical work of digital humanitarians—teams of everyday people across the globe who curate social media and other online information for crisis responders on-the-ground during a mass emergency. Specif- ically, this work identifies and investigates a new concept of breakdown, coined sociotemporal disorder, across empirical, theoretical, and design constructs to understand how time is socially structured and experienced in humanitarian work; the temporal sensemaking strategies the team uses to discern information; and new ways to reveal breakdowns in time as a resource for the design of sociotechnical tools appropriated for this work. Together, this research contributes a roadmap to address a longstanding gap in the CSCW and HCI literature by interleaving the so- cial experience of temporality with instrumented time to mitigate against sociotemporal disor- der to allow distributed online work to productively unfold. This research, as a whole, foretells future work in exploring sociotemporal disorder as a resource in the design of sociotechnical sys- tems. iii

Acknowledgements

I am grateful beyond words to the following people who generously guided me through the doc- toral program and helped me blossom as a scholar and teacher.

To my advisor Stephen Voida, a kind old soul whose brilliance is without peer. Thank you for your patience and friendship on this incredible journey.

To my indefatigable co-advisor, Leysia Palen, who inspires me daily to follow my dreams and to embrace small comforts in flowers and dogs when the humanitarian work breaks my heart.

I am equally indebted to my committee Amy Voida, Brian Keegan, and Katrina Petersen. Thank you so much for your willingness to lend support and advice when it mattered most.

To Janghee Cho and Lucy Van Kleunen in the Too Much Information Lab and all the generous souls in the CU Boulder Information Science graduate cohort who prove daily that a caring cul- ture is a wellspring for the best scholarship.

To my daughter, Drea Norris, whose thoughtfulness and resilience no know bounds. I am so in- credibly proud of you.

To Howard and Terry Norris, my father and step-mother, who both passed away during my stud- ies. You instilled in me the value of hard work, a love of reading, and the importance of staying grounded. Rest in power. I miss you both very much.

In closing, I dedicate this thesis to my digital humanitarian colleagues at The Standby Task Force—

Valeria, Joyce, Leisa, Deepshikha, Stuart, Per, Liz, and the thousands of SBTF volunteers from

106 countries. It has been my supreme honor to join together with all of you to help ease suffer- ing in the world. iv

Contents

Chapter

1 Introduction 1

1.1 Background ...... 3

1.1.1 Information and Communication Technologies ...... 3

1.1.2 Humanitarian Crisis and ICTs ...... 4

1.1.3 Crisis Informatics and ICTs ...... 4

1.1.4 Digital Humanitarians ...... 5

1.1.5 The Standby Task Force ...... 6

1.2 Research Questions ...... 7

1.3 Overview of the Research ...... 9

1.4 Style of this Dissertation ...... 14

1.4.1 Chapter 2: Is the time right now? ...... 14

1.4.2 Chapter 3: How to make sense of time? ...... 15

1.4.3 Chapter 4: Designing for sociotemporal disorder ...... 16

1.4.4 Chapter 5: Discussion ...... 16

1.4.5 Chapter 6: Conclusion and Future Work ...... 16

2 Is the Time Right Now? Reconciling Sociotemporal Disorder

in Distributed Team Work 17

2.1 An Unprecedented Disaster ...... 17 v

2.2 Overview and Contributions of this Research ...... 19

2.3 Temporality and Work ...... 20

2.4 Digital Humanitarians ...... 24

2.5 Methods ...... 26

2.5.1 Research Context ...... 26

2.5.2 Corpus ...... 30

2.5.3 Data Analysis ...... 30

2.6 Pluritemporal Analytic Framework ...... 33

2.6.1 Three Temporal Orders ...... 44

2.6.2 Temporal Coordination: The Work of Reconciling Sociotemporal Disorder . . 49

2.7 Discussion: Designing to Support Temporal Coordination Work ...... 51

2.8 Conclusion ...... 57

3 Narrating Time as Temporal Sensemaking 59

3.1 Introduction ...... 59

3.2 Background: Digital humanitarian work ...... 60

3.2.1 Case: 2017 Hurricane Maria ...... 61

3.3 Related Work ...... 62

3.3.1 Social Constructions of Time ...... 62

3.3.2 Temporal & Distributed Coordination ...... 62

3.3.3 Temporal Sensemaking ...... 63

3.3.4 Temporal Narrative ...... 63

3.4 Methods ...... 64

3.4.1 Participants & Recruitment ...... 65

3.4.2 Data Collection ...... 65

3.4.3 Data Analysis ...... 66

3.5 FINDINGS ...... 66 vi

3.5.1 Triaging Temporal Information ...... 67

3.5.2 Evaluating Temporal Information ...... 69

3.5.3 Negotiating Breakdowns in Temporal Information ...... 70

3.5.4 Synchronizing Temporal Information ...... 72

3.5.5 Crafting the Emergent Temporal Narrative ...... 74

3.6 Speculations on new temporal sensemaking approaches ...... 75

3.6.1 Triage Tools to Scale Good Will ...... 76

3.6.2 Visual Representations of Time & Evaluative Metadata ...... 77

3.6.3 Negotiate Clustered Information Rather than One-offs ...... 77

3.6.4 Synchronization Across the Team through Feedback ...... 78

3.7 Discussion and Implications ...... 79

3.8 Conclusion and Future Work ...... 80

4 Designing for Sociotemporal Disorder 82

4.1 Introduction ...... 82

4.2 Digital humanitarians as a community of practice ...... 84

4.3 Applying participatory design to sociotemporal disorder ...... 86

4.4 Methods ...... 87

4.4.1 Overview of the virtual PD approach ...... 87

4.4.2 Participants and recruitment ...... 88

4.4.3 A novel adaption of participatory design for virtual teams ...... 89

4.4.4 Temporal priming of workshop teams ...... 91

4.4.5 Data corpus ...... 92

4.4.6 Data analysis ...... 93

4.5 Findings ...... 93

4.5.1 Roadblocks to distributed temporal work ...... 93

4.5.2 Roadblock analysis ...... 94 vii

4.5.3 Determining design priorities ...... 97

4.5.4 Solutions analysis ...... 100

4.6 Discussion ...... 104

4.6.1 Themes across artifacts ...... 105

4.7 Design Implications and Future Work ...... 107

4.8 Conclusion ...... 108

5 Designing for Sociotemporal Disorder 110

5.0.1 New interpretations of speculative designs ...... 113

Bibliography 115

Appendix

A SBTF vPD workshop protocol 126 viii

Tables

Table

4.1 Study participants by SBTF organizational role and team size ...... 90

4.2 Activity protocol for each vPD workshop...... 109

A.1 Activities and Scripts ...... 126 ix

Figures

Figure

1.1 Overarching research ...... 10

1.2 ...... 11

1.3 ...... 12

1.4 ...... 13

2.1 Hurricane Maria crisis map ...... 27

2.2 Examples of crisis tweets ...... 29

2.3 SBTF Google Sheet for crisis data collection ...... 29

2.4 SBTF Slack thread ...... 31

2.5 Facebook post representing time as an interval ...... 35

2.6 Tweet representing time as chronology ...... 37

2.7 Example of conflicting data on SBTF Google Sheet ...... 44

2.8 Facebook post representing time as implicit ...... 46

2.9 Tweets representing locational timezone metadata ...... 46

2.10 Qualitative representation of time ...... 48

4.2 Roadblocks affinity map ...... 94

4.3 Roadblocks spectrum diagram ...... 95

4.4 Priority affinity map ...... 97

4.5 Design sketches ...... 100 x

4.6 Solutions affinity diagram ...... 102

4.7 Solutions spectrum diagram ...... 102

5.1 Tweetdeck prototype design ...... 114

5.2 Heat map prototype design ...... 114 Chapter 1

Introduction

“What time is it?” is a deceptively simple question. Temporality, or the social organization of time, is an essential element of human experience—it shapes our work practices, social rhythms, and the interpretative frames we use to interact with and make sense of the world [149]. Clocks, timestamps, time zones, and temporal artifacts, like transportation schedules, enable people to share a mutual understanding of time [115, 149]. Establishing a common temporal order allows people to orient themselves around a collective agreement about “what time it is.” Once oriented to a shared clock-time, we use sociotemporal orders as a common framework to put time into practice and ascribe collective meaning to it. The synchronization that results from sociotem- poral order occurs across both large and small groups via basic technologies, such as the town center clock tower that coordinates community activities in the moment [51] and to more com- plex processes, like critical workplace decision-making conducted in due time [85].

Temporality is often taken for granted until there is a coordination breakdown. As modern life and work are increasingly technically mediated by ubiquitous devices, like mobile phones that reference a single time source, the precision of clock-time influences temporal coordination in new ways [74]. However, there are discrepancies between mechanical representations of time and the human experience of sociotemporal order. For instance, people socially construct time in a variety of ways, as experiences, events, etc., while information and communication technologies

(ICTs) that ostensibly allow people to share moments in time also tend to represent time as fun- gible, standardized units [136]. The rise of ubiquitous and cloud-based collaborative ICTs that 2 enable geographically dispersed people across multiple time zones complicates the capacity to temporally coordinate work; where the question “what time is it?” legitimately has multiple an- swers.

In my research, I am drawn to understanding one of the more enduring challenges in computer- supported cooperative work (CSCW): how to better support online collaboration in times of global crisis to ease human suffering and loss. There are two fundamental issues at play. One, inflexible sociotechnical systems that are designed to solve social problems and enable well-being but are still largely attuned to reflect time as mechanical. This results in a heavy cognitive bur- den for complex social encounters, such as coordinating globally-dispersed crisis response in time- and safety-critical situations. Two, temporal experiences are incredibly complex social phenom- ena and difficult to deconstruct in online spaces. Social constructions of time take several forms, as shared interactions and turn taking [71], cross- cultural notions of time [88], social rhythms

[127], routines [149], frames of reference [71], ephemerality [56], pluritemporality [98], temporal depth [10], contextual timescapes [4], and temporal processes [113].

Accordingly, this research aims to better understand: how do globally distributed teams temporally coordinate their computer-supported cooperative work?

My dissertation research builds on interdisciplinary literature about time, its various social rep- resentations, the ways that temporal experiences are ordered and disordered, and how temporal sensemaking is conducted in real-time crisis scenarios. The aim of this dissertation is to deepen theoretical and empirical scholarship about how sociotemporal order is established to allow dis- tributed temporal coordination work to productively unfold and to reveal pathways to improve the design of technically-mediated CSCW to help reconcile sociotemporal disorder.

The research that follows explores an extreme case [103]: global digital humanitarian work [132,

28, 92, 93] in the aftermath of the 2017 Hurricane Maria catastrophe in Puerto Rico. The high- tempo, time- and safety-critical nature of this virtual information work exposes the crucial role of temporal coordination in crisis response and amplifies the breakdowns that threaten the in- tegrity of the work. Further, this domain opens new avenues to explore how temporality emerges 3 as qualitative experiences of time in work, technologies, and artifacts.

Time and temporality are often cited as important elements in digital humanitarian work for sensemaking [102], decision support [52, 135], information sharing [29], and information qual- ity [76]. Crawford and Finn [34] have also called for critical data studies to consider the visible and invisible influences of temporality on disaster response and the ways that technologies can recursively shape crisis content. The temporal concerns raised here are consistent with my own

five years of experience as an active volunteer on a digital humanitarian team. My participant- researcher status helps to inform and instantiate this work.

1.1 Background

In the following section, I introduce several related concepts that are key to understanding the complex coordination work of digital humanitarians and the inherent sociotechnical challenges these groups face.

1.1.1 Information and Communication Technologies

Since the early 2000s, ICTs allow people to come together online to collaborate, socialize, and share [49]. Over the past two decades, mobile devices have become ubiquitous across the world coupled with rapid growth in internet access [64]. The advent of social media and free, cloud- based collaborative tools, like document sharing, messaging, and open-source mapping, have made a lasting impact on how humans interact and coordinate activities in real time, particularly during moments of crisis.

Posing the question “what time is it?” is a useful probe to interrogate the temporal features and affordances of ICTs used by the digital humanitarian team, just as it is for exploring how they conduct their distributed temporal coordination work. Here, the aim is to reveal the relative and relational ways that time is technically mediated as temporal representations, orders, disorders, and discourse in real time as well as the implications for effective coordination that produces timely, trustworthy information artifacts. Key areas of focus in the design work of this research 4 are: 1) the imposed temporal ordering of social technologies, e.g., the algorithmic illusion of “real time” that represents a false sense of immediacy [94]; 2) the lack of flexibility in representing dif- ferent modalities of time, e.g., pluritemporal discourse [93]; 3) invisibility of useful temporal in- formation embedded in metadata, e.g., missing API data in consumer-oriented social media apps

[94]; 4) the dearth of temporal cues, time perspectives, and polychronic preferences about the on- going coordination work, e.g., inability to convey a “collective mind” about task awareness [96].

1.1.2 Humanitarian Crisis and ICTs

A humanitarian crisis is triggered by an environmental hazard, public health threat, and/or con-

flict that causes mass casualties, sudden displacement of people, disruption to civil society, and wide-spread property loss [109, 91]. ICTs have played a key role in helping people communicate during sudden catastrophes [84, 92]. During a crisis, people often use social media to share infor- mation, document eyewitness reports, and even call out for help when public emergency commu- nication systems fail. In the case of social media, tens of millions of data points can be generated due to a single large-scale humanitarian crisis. The resulting data deluge alongside the availabil- ity of free, collaborative ICT tools has provoked a sea-change in global crisis response coordina- tion, information gathering, and crisis data analysis [92, 94].

1.1.3 Crisis Informatics and ICTs

The use of ICTs during a crisis—to communicate about local conditions and to collect data to inform the response—has given rise to an emerging subfield of information science. Crisis in- formatics [101] integrates human- and machine-computation, sociotechnical systems, and hu- man–computer interaction (HCI) methods to improve information sharing about crisis warning, response, and recovery. To date, crisis informatics research has largely been driven by the ca- pacity to collect large amounts of digital data from social media, construct online maps, and ex- tract novel insights from metadata and contextualized social media posts. While this approach is ideal for studying the intersection of data and behavior at large scale, Palen and Anderson 5

[101] contend that more precisely scoped information can also offer important qualitative, ethno- graphic insights about decision-making contexts, collaborative behaviors, and temporal discourse.

A grand research challenge for crisis informatics over the next decade is the study of social con- structions of time and how these human factors should be better represented, integrated, and supported in humanitarian-motivated crowd work [95].

1.1.4 Digital Humanitarians

Ad hoc distributed teams of global volunteers first came to notice in the wake of the catastrophic

2010 Haiti Earthquake by which estimated 160,000 people perished and 3 million more affected

[131]. With the advent of social media, cloud collaboration and geospatial mapping tools, and camera-enabled mobile phones, people began to form virtual networks of “everyday analysts”

[102] to curate local and timely information about distant emergencies [129, 132, 28, 61, 125, 76,

58, 92].

Dubbed digital humanitarians [84], globally distributed online volunteer teams provide rapid crowdsourcing of time-critical information, media monitoring, data analysis and verification, satellite imagery analysis/mapping, language translation, and software development. This infor- mation support can ramp up within the first few hours of a crisis to provide initial situational awareness about local conditions as international crisis response organizations race to mobilize people and resources. Digital humanitarian groups refer to their distributed online work as an ac- tivation to signal the call to action by their formal partners, as well as to differentiate it from an on-the-ground deployment of emergency responders.

Digital humanitarian teams formalized their operations and protocols, adopted codes of conduct, and honed work specializations. Soon after, the leading crisis response institutions, including the

United Nations Office for the Coordination of Humanitarian Affairs (UN OCHA), World Health

Organization, M´edecinsSans Fronti`eres(MSF), and the U.S. Federal Emergency Management

Agency (FEMA), among others, initiated partnerships with digital humanitarian teams (The

Standby Task Force, n.d.). 6

This work, however, is not without its detractors. Prior work in critical and feminist geographic information science, social science, and law/ethics have raised important concerns about the col- lection, processing, and use of humanitarian crisis information, particularly those derived from public “big data.” These massive datasets of social media posts require sophisticated processing and careful social interpretation. An overreliance on the technical hype surrounding big data can create unintended, and possibly harmful consequences for crisis-affected people. Concerns about power dynamics, social inequalities, privacy and consent, and technology access demand atten- tion in crisis informatics research [18, 34, 77, 81]. In particular, the collision between arms-length

(big data) sense making and cultural knowledge (local expertise) reifies social problems around power and agency [111]. Ultimately, these unresolved concerns can lead to ill-informed and inef- fective, if not dangerous, crisis response. These salient critiques come into sharper focus in my decision-making about addressing power dynamics in the research design of the study described in Chapter 4.

1.1.5 The Standby Task Force

My research partner is The Standby Task Force (SBTF), one of the first digital humanitarian teams to emerge from the 2010 Haiti earthquake response. Currently staffed by 2,100 trained volunteers from 106 countries, SBTF is widely regarded for its capacity to deliver 24/7, rapid- response distributed work and information crowdsourcing expertise.

Over the last 10 years, the SBTF has collaborated with leading global crisis response institutions to collect, verify, and synthesize public data from social media, news, official communiques, and satellite/UAV imagery to produce situational awareness, remote damage assessments, and crisis maps for use by emergency responders in the field [92, 93]. SBTF provides its services at no cost and, like many nonprofit organizations, appropriates assemblages of information systems to carry out their work [114, 140]. SBTF uses free, cloud-based systems in three contexts: (1) as source material for their work, e.g., volunteers comb through public websites and social media feeds to identify relevant information about crisis-affected people; (2) to create and distribute new infor- 7 mation artifacts via Google Office Suite tools and ArcGIS; and (3) to coordinate their work, e.g., through the Slack chat app.

SBTF has kindly agreed to participate as an ethnographic site, provided access to its Hurricane

Maria document archive, and granted access to SBTF volunteer logs to recruit study partici- pants.

1.2 Research Questions

Despite more than 10 years of partnership with the crisis response sector, there has been surpris- ingly little direct attention in the literature on exploring how time is implicated in the work prac- tices of digital humanitarian teams [28, 34, 95, 132]. Thus, this research is framed by an overar- ching question: how do globally distributed teams temporally coordinate to accomplish their work?

In order to reduce the abstruse nature of temporal work coordination in a crisis domain, my first set of research inquiries are designed to unpack how the process of sociotemporal ordering is con- structed. Drawing from Lincoln and Guba [72], these questions are designed to triangulate within and across each way the team encounters time and how time is reflected back to them through data.

The idea is that time is relative to each member of the team, to each piece of disaster-related in- formation, and to the various collaborative information systems that each differently encode and display time. By interrogating the historical chat dialogue during the Hurricane Maria activa- tion and other temporal work artifacts, as well as contemporary reflections about the completed project, time becomes more concrete. Here, the theoretical process of achieving temporal coordi- nation is synthesized into three component parts that form the initial empirical inquiry in Chap- ter 2, a reprint of Norris, Video, Palen, and Voida [93]:

• RQ1a: How do globally distributed teams establish shared understandings about time?

• RQ1b: What is the role of discourse in temporal coordination? 8

• RQ1c: How do globally distributed teams achieve shared temporal understandings?

How do they reconcile breakdowns in agreement?

Where RQ1 unpacks how temporal order is achieved and acted upon by the team, the next step in the temporal coordination process engages how the team constructs a collective temporal un- derstanding about ground conditions from different forms of time that arise from the crisis and the technology they use in their work. For the SBTF volunteers, this means collectively interact- ing with public social media data, applying and synthesizing multiple interpretations, and then shaping it into actionable, time-critical information for the crisis response activation partner.

Due to the nature of the extreme high-tempo information flows, this form of distributed coordi- nation is intensely dialogic [46], relying on constant communication, situational improvisation, and the collective perspectives of the team toward joint sensemaking. In this context, sensemak- ing applies an interpretive lens to collectively construct understanding [107, 146]. However, sense- making in this domain also has a temporal component to it since the goal of the Hurricane Maria activation is to assemble local situational awareness in as near a real-time chronology as possible.

The sensemaking artifact, be it a crisis map or data report, is a form of temporal narrative—a series of collectively negotiated interpretations of an event over multiple points of view, contexts, and moments in time [37]. Here, the concept of sociotemporal orders also plays an important role in constructing a cohesive temporal narrative. The “story” of the data now has a shared social context for further interpretation and decision-making by FEMA. How this temporal sense- mak- ing and narrative process actually unfolds across the distributed team, however, is not well un- derstood. Chapter 3 is comprised of a standalone manuscript for a second study which asks:

• RQ2a: What temporal sensemaking approaches do globally distributed teams use in

high-tempo, time- and safety-critical information work?

• RQ2b: How do globally distributed teams collectively construct a temporal narrative

from its sensemaking process to convey a coherent data story? 9

The final set of research questions are designed to put the empirical findings (R1 and R2) about coordination practices, pluritemporal discourse, sociotemporal order/disorder, and temporal sense- making into practice. Here, I propose the co-design of new tools and/or work practices in col- laboration with the SBTF team. Elaboration of the design protocol to address R3 and R4 is de- scribed in detail in a recently completed study in Chapter 4.

• RQ3a: How might globally distributed teams reconcile the sociotechnical breakdowns en-

tangled between distributed temporal coordination work and appropriated collaborative

tools?

• RQ3b: In what ways might collaborative technologies better support the temporal coordi-

nation work of globally distributed teams?

• RQ4a: How can globally distributed teams productively participate in a virtual co-design

process?

1.3 Overview of the Research

Time and temporality as topics of inquiry easily slip into high-level abstractions that are quite difficult for people to conceptualize, concretize, and articulate. Initial pilot interviews with the

SBTF leadership team about this research confirmed early on how hard it is to talk about some- thing that feels so intangible. To mitigate these problems in my dissertation research, I have taken a multimethod approach that allowed me to interrogate the research questions from var- ious perspectives, e.g., human:human and human:computer interaction, etc., and from different data types, e.g., discursive, observational, metadata, etc. Specifically, the research (Table 1) used a combination of ethnographic study, semi-structured interviews, document analysis, qualitative coding, participatory design, and feedback surveys. Further, each of the three studies (elaborated below) were designed to produce trustworthy results through prolonged engagement, established qualitative procedures, reflexive memo writing, and triangulation of methods and data [72, 122]. 10

Figure 1.1: Three research studies designed to explore distributed temporal coordination in CSCW and HCI. 11

Figure 1.2: Temporal Sensemaking Study. 12

Figure 1.3: Designing for Sociotemporal Disorder Study. 13

Figure 1.4: Virtual Participatory Design Method. 14

The overarching research is grounded in more than four years of digital ethnographic study of

SBTF and its globally distributed volunteer corps, as well as insights from my own participation as an SBTF volunteer in 14 disaster activations since 2015. Longitudinal ethnographic work is fairly atypical in CSCW research. However, the benefit of nearly five years of close engagement with the team revealed novel pathways to explore their distributed temporal coordination prac- tices, technology appropriation, and the temporal breakdowns inherent in broadly distributed work. Further, the long-standing relationships with the SBTF leadership team ensured unfettered access to activation archives. However, there are cautions with participant- research that require deep reflection and mitigation. In response, I engaged in a bracketing process of extensive memo writing and debriefs with my advisor and lab colleagues to draw awareness to and then set aside preconceptions about emerging insights from the research [5, 138].

1.4 Style of this Dissertation

Through a combination of empirical, theoretical, and design studies, Chapters 2-4 are presented as cohesive units of scholarship that successively build a research narrative about sociotemporal dis/order and the sociotechnical systems needed to support distributed temporal coordination work. This topic has received surprisingly little direct attention in the CSCW and HCI literature even as geographically-dispersed remote collaboration continues to gain ground in the workplace.

1.4.1 Chapter 2: Is the time right now?

The central question in the first study is “How does the globally distributed SBTF digital human- itarian team temporally coordinate its work?” To address RQ1 and RQ3, we interrogated Hurri- cane Maria Slack chat logs and associated artifacts, including the Google Sheet used for crowd- sourced data collection, the ArcGIS map populated from the Google Sheet data, and instruc- tional activation documents used to “onboard” the volunteers. This theoretical-empirical study presents four contributions: 1) A pluritemporal analytical framework to understand how to dis- ambiguate different meanings of time in the Slack chat discourse; 2) empirical evidence of how 15 the team establishes collective agreement about time to enable productive work (sociotemporal order); 3) a description of coordination breakdowns caused by an inability to reconcile agree- ment about time (sociotemporal disorder); and 4) reflections on the design implications for ap- propriated collaborative technologies that may help to reconcile sociotemporal disorder in dis- tributed coordination work online. This study also reveals that temporal language in the chat logs is highly relative to its situational context. Further, understandings of time are quite cul- turally entangled through the use of local metaphors, the nature of work and time- keeping, and experiences shaped by technology [3, 4], Lindley2015. This is also the case in verbal discourse, which makes discussion about time and temporality quite difficult for the globally distributed

SBTF volunteers to talk about in direct terms. This research was presented at CSCW 2019 and published in Proceedings of the ACM. It is reprinted in full as Chapter 2.

1.4.2 Chapter 3: How to make sense of time?

Building on insights from the document analysis study, my long-term ethnographic work with

SBTF, and my own experience during activations, the second study focuses on temporal sense- making and the strategies that digital humanitarian volunteers use to construct a coherent tem- poral story. To address RQ2 and RQ3, I center semi-structured interview questions posed to nine

SBTF volunteers who participated in the Hurricane Maria activation. Using chat excerpts and other artifacts from Study 1 as probes allowed me to concretize the notion of temporality with which to ground the interview dialogue. This completed research contributes a descriptive ac- count of four collective temporal sensemaking approaches (triaging, evaluating, negotiating, and synchronizing) that the SBTF team uses to construct a coherent data narrative about the crisis zone. I also posed “magic wand” questions to the participants to prompt speculative design ideas that would better support temporal coordination practices and technologies. 16

1.4.3 Chapter 4: Designing for sociotemporal disorder

The final study uses an applied research approach to integrate the contributions of the previous studies to explore HCI-focused concerns via RQ3 and RQ4. Specifically, this study brings to- gether a cross-section of SBTF volunteers to explore how they understand and engage with (or ignore) the multitude of time-related issues that often invisibly shape their ad hoc work practices and the appropriated sociotechnical systems they must use to conduct their work. Here, we use a novel virtual participatory design (vPD) protocol to guide small volunteer teams through on- line workshops to collectively identify roadblocks, prioritize needs, and then envision potential solutions grounded in design justice values. The vPD protocol was deliberately crafted to pro- vide a useful co-design template for pre-existing online teams that also provokes opportunities to confront the power dynamics of new configurations of labor online and the free, commercial technologies used in service of their digital humanitarian work. This research contributes several empirical insights about how sociotemporal disorder manifests in surprising ways in distributed temporal coordination work, as well as practical implications to guide the design of cloud-based sociotechnical tools to propel the future of remote work in useful and more humane ways.

1.4.4 Chapter 5: Discussion

Here, I present a summary and discussion of the findings across the three studies. In addition to a comprehensive discussion of the implications of this work for the CSCW and HCI communities,

I present a set of initial design interpretations that will be used as discussion probes for future work with SBTF exploring sociotemporal disorder as a resource for design, instead of a break- down for designers to reconcile.

1.4.5 Chapter 6: Conclusion and Future Work

In conclusion, Chapter 6 reiterates the contributions of this dissertation and foregrounds a fu- ture research agenda that advances the design of collaborative sociotechnical systems as dynamic socially intelligent computing that is responsive to sociotemporal experiences. Chapter 2

Is the Time Right Now? Reconciling Sociotemporal Disorder in Distributed Team Work

2.1 An Unprecedented Disaster

In the aftermath of the catastrophic destruction of Puerto Rico by Hurricane Maria in 2017, volunteer digital humanitarians [84] from around the world came together online to help. At the request of the U.S. Federal Emergency Management Agency (FEMA), volunteers with The

Standby Task Force (SBTF)1 curated time- and safety-critical information about the highly fluid operational status of hospitals across the island. This type of globally distributed work presents multiple coordination challenges that arise from the almost invisible combination of temporal or- ders, or shared understandings of time within a group. Collective assumptions about time are characterized by shared rhythms, patterns, and reference points to which individuals orient their own experiences—in the natural world, through technology, and in society. The different repre- sentations of time that help people establish a shared sense of temporal order are quite diverse.

For example, we all agree that midday can be represented by the regular diurnal cycles caused by the Earth’s rotation, when the clock strikes 12 p.m., as a synonym for “lunchtime,” and as the timestamp 2019-11-09T12:00:00Z.

Here, this shared understanding of clock time leads to a sociotemporal order, a common frame- work for putting time into practice, which establishes that “midday” relates to the timing and context around a phenomenon in which both collective agreement and organization is needed in

1https://www.standbytaskforce.org 18 order to do something, like taking a break from work, eating lunch, or—as in our data—notating an event with a high level of precision.

These challenges are in full view in the following verbatim coordination chat transcript excerpt 2 between two SBTF volunteers:

Leia [2017-09-28T09:46:06Z]: I just made sure all the hospitals the [Depart- ment of Health] says are up and running as of yesterday had at least 1 report that they were open. It certainly is confusing as there are social media reports seemingly from the same time saying they are not open.

Leia [2017-09-28T09:46:22Z]: Hi Lexi, good morning, or almost afternoon here.

Lexi [2017-09-28T09:48:11Z]: Yeah I found a lot of conflicting information last night. I think the best we can do is just get it all down and see if we can make some sense of it prioritising most reliable sources and time stamps. Thank you!

In the first turn-take, Leia engages with no fewer than three forms of time. First, she declares work she is doing relative to her own sense of time (“just”, implicitly leaning on the chat plat- form timestamp to resolve any potential ambiguity about when she performed the work). Second,

Leia refers to “yesterday” as a chronology of her work. Third, she refers to metadata from social media posted “seemingly from the same time” that she is using to derive information.

Leia then welcomes Lexi to the group chat. In her greeting, Leia establishes the local time as “al- most afternoon” for both team members who reside one time zone apart.

Lexi responds in the last turn-take with three different time references after “yeah” to perhaps confirm the temporal agreement of the relative nature of their work. She refers to “last night” to situate the solar rhythm of her most recent work in contrast to re-entering the chat room now.

Lexi describes “prioritising” as a time-relative strategy to evaluate the information the group has collected, thus far. She then, too, mentions metadata “time stamps”—drawing attention to how

2Volunteer names are anonymized for privacy, and the sociotemporal discourse coded in our analysis is dis- played in italicized text for clarity. 19 technology affects the product of their work: the creation of a chronology of available hospital resources to hurricane victims.

It is this wrestling with different temporalities—natural, social, and technological—that appears in the sometimes-convoluted discourse that this paper unpacks, theoretically describes, and then considers as grist for technology design for distributed, computer-supported collaborative work.

2.2 Overview and Contributions of this Research

Temporal coordination is fundamental to the functioning of society. Shared orientations to time shape rhythms, patterns of coordination, and interpretive frames for making our way in the world

[149]. The creation of these shared orientations and the construction of a shared sense of “what time it is” [115, 149] is a form of work—temporal coordination work. However, temporal coor- dination endures as a central topic in computer-supported cooperative work (CSCW) because information systems still struggle to adequately support varying representations of time in the context of collaborations that are both temporally and geographically dispersed.

We present empirical evidence of how SBTF members use a variety of discursive temporal refer- ences to overcome instances of sociotemporal “disorder”—breakdowns in coordination caused by the inability to reconcile multiple temporal orders that they draw upon in the conduct of their work. For these digital humanitarians, temporal coordination is hampered by the chaos of incom- ing online information streams, varied metadata structures, widely dispersed workforce across numerous time zones, and inflexible technical tools that cannot readily handle other forms of temporal representations aside from clock time, e.g., shared interactions and turn-taking [71], cross-cultural notions of time [88], and qualitative experiences of temporality [113]. An analysis of the work of SBTF reveals that this temporal coordination work is complicated by the need to reconcile multiple different temporal orders—what time it is relative to one another, to each piece of disaster-related information, and to the various collaborative information systems that each differently encode and display time.

Here, we interrogate an extreme case of temporal coordination, a case in which the phenomenon 20 of study is intense [103]: high-tempo digital humanitarian information work in the aftermath of the 2017 Hurricane Maria, one of the deadliest Atlantic hurricanes on record. In this context, the time-critical nature of the work further exposes and amplifies how temporal coordination must be achieved. Digital humanitarian workers—often volunteers [28, 92, 132]—are activated from around the globe as time and expertise allow. People join in anew with each event, putting fur- ther demands on the work that must be done to achieve temporal coordination.

Through an analysis of this case that foregrounds a nearly invisible substrate of the temporal coordination of distributed work, we make four contributions:

A pluritemporal analytic framework to interrogate the forms of language that people employ to resolve their sociotemporal disorder; A descriptive account of the different temporal orders that our informants must reconcile, which was revealed by our application of the analytic framework;

A theoretical description of how sociotemporal disorders are created and must be reconciled at the intersection of multiple, conflicting, temporal orders, including sociotemporal orders; and A suite of design implications for collaborative information systems to better support temporal co- ordination work.

There has been surprisingly little direct attention in the literature on exploring the temporal co- ordination practices of globally distributed digital humanitarian groups [28, 92] or the sociotem- poral orders that influence their work [132], despite more than 10 years of partnership with the emergency response sector. Further, this ethnographic work helps to reveal a pathway toward un- derstanding the larger and more complex practice of technology appropriation in CSCW [133].

This study is designed to help address these gaps.

2.3 Temporality and Work

Work—from interpersonal to institutional—depends on coordination with others [120], includ- ing temporal coordination. Technologies have been invented for the purposes of synchronization, such as the town center church clock to coordinate geographical communities [51] and the wrist- watch to coordinate war maneuvers by people distributed over vast areas [80]. Temporal artifacts 21 and new normative behaviors have in turn arisen because of technological innovation. The rail- road gave rise to coordinated clock time across towns and cities, as well as to time zones across regions [150]. The precision of clock time on mobile phones, which reference essentially a single time source, means that “being late” is not a function of individually calibrated wristwatches, and therefore bears on temporal coordination behavior in new ways [74].

Temporal coordination relies on a shared knowledge of “what time it is” (and what day and date it is) [149], as well as the capacity of people to act on that knowledge in a synchronized way [8], in due time [85], or across locally distributed groups [25]. We see this shared temporal knowl- edge in action when a small group imposes local meaning (e.g., the video conference is always at

2:00pm), a large group imposes structure on the year (e.g., a national holiday), or the community indirectly imposes the consequences of their coordinated habits (e.g., leaving work around the same time of day, which creates the traffic rush) [10, 56].

Recurrent collective activities give rise to the need for social orders, or a set of structures, sys- tems, values, and norms that guide interactions. When a social order is coupled with shared temporal knowledge (“what time it is”), the resulting sociotemporal order [149] provides a com- mon framework about how we think about time, ascribe meaning to it, and put shared temporal knowledge into practice. The seven-day week of a Gregorian calendar is an example of a conven- tional sociotemporal order that regulates social life. Sociotemporal orders are further elaborated by localized rhythms that are particular to smaller social groups (of the household, of religious groups, of a workplace’s shift patterns, etc.) [149].

Sociotemporal orders are regulated by quantitative measures, typically clock-time or diurnal cy- cles, and qualitative features, such as social rhythms, patterns, routines, frames of reference, ephemeral states, and events [1, 51, 56, 71, 73, 88, 91, 127, 142, 149]. Sorokin and Merton [127, p. 623] describe features of qualitative time as “derived from the beliefs and customs common to the group and ... serve further to reveal the rhythms, pulsations, and beats of society in which they are found.” Shared understandings of qualitative representations of time are relative, which can lead to confusion and discord—especially in diverse, globally distributed groups. Yet, these 22 qualitative features are sometimes the only consistent or reliable sociotemporal information avail- able to the group.

Grosz [53] describes how events can act as temporal ruptures of the past that transform the present and future in unpredictable ways. Here, Hurricane Maria caused a catastrophic rupture to the rhythms of life in Puerto Rico and elsewhere in the eastern Caribbean where the hurricane made landfall. The sociotemporal disordering of social institutions, infrastructures, and personal lives are experienced not only in the timing of disrupted daily routines according to the clock but the qualitative aspects of urgency, priority, and change in responding to the immediate post- hurricane reality. The resulting turmoil may play out for years to come.

Sociologists who study time have written about how historical events, theologies, technologies, and political systems have all shaped social constructions of time over thousands of years [2, 50,

71, 98, 127, 142, 151]. In this research, we examine how temporal coordination is negotiated in

CSCW (see also [44, 73, 82, 112], and geographically distributed CSCW, more specifically (see also ([100]). The ubiquity of distributed teams is on the rise, made possible by (often) free soft- ware that enables synchronous and asynchronous collaboration through video conferencing, real- time messaging, collaborative text editing, and shared information repositories. Yet, distributed work is a context in which we understand significantly less about how teams coordinate tempo- rally to accomplish their work. In the meantime, we look to adjacent high-tempo, information- centric coordination work for insights. Computer-supported temporal coordination in hospitals evokes a complex interleaving of quantitative and qualitative sociotemporal representations, in- cluding clock-time, scheduling, synchronization, and temporal allocation, in both patient care and staffing [8]. Moreover, temporal coordination requires significant support—whether through precise timekeeping in the case of a resuscitation team [67] or deep understanding of how tem- poral experiences influence work activities [112]. Here, Reddy et al identify three qualitative so- ciotemporal features—trajectories, rhythms, and horizons—that contextualize and drive the co- ordination activities of complex information work across temporally-distributed staff on different shifts in a surgical intensive care unit (SICU). 23

These sociotemporal features reflect how courses of action over time, patterns of activities in the unit, and future events are used by care teams to negotiate the subjective temporal relations nec- essary for effective collaboration. This framework, in particular, stands out for its thick descrip- tion of the tensions of integrating qualitative temporal information into practice which is quite analogous to digital humanitarian work. However, the place-based nature of the coordination work by stable, professional teams makes it difficult to draw adequate comparisons for ad hoc, globally distributed volunteers with more constraints and fewer resources to achieve temporal awareness.

Recent work on the mediating role of technology on temporal coordination suggests that when people encounter multiple and recurring demands on their time, they adopt different approaches to enact a sense of sociotemporal order. Erickson and Mazmanian introduce the concept of tem- poral entrepreneurship to explore how some knowledge workers are pushing back against estab- lished norms of time. Here, they describe innovative ways people are using mobile technologies to characterize new qualitative forms of sociotemporal orders, such as: spectral time which de- scribes how imprecise trajectories (borrowing the temporal arc concept from Reddy et al [112] are transformed to micro-coordinate brief, iterative activities; cohabited time that integrates mul- tiple physical and distributed collaboration ecosystems of work and life; and porous time that simultaneously overarches multiple social contexts at once [44]. These entrepreneurial practices are especially relevant to digital humanitarian work as SBTF volunteers also encounter multi- ple sociotemporal orders—in the crisis information that they collect, technologies that they use, and distributed social coordination processes they engage in (see the next section for elabora- tion on these points). Likewise, SBTF volunteers attempt, with varying degrees of success, to innovate coordination practices and appropriate technologies as workarounds to sociotemporal breakdowns, though in a fast-paced, highly distributed, and large group context, in contrast to individual temporal entrepreneurs.

With so many different and simultaneous temporal factors at play, we look to the concept of pluritemporality [98] to explore the different ways that the SBTF volunteers talk about time to 24 help reconcile sociotemporal disorder in their information work. Pluritemporality is a social con- struct that describes how people interpret and assign meaning to modes of time(s) which may exist side-by-side [98]. In this sense, adopting a pluritemporal framework helps to reveal the nearly invisible substrate of temporal language and representations that are necessary for shared understandings of time, and which ultimately help to support temporal coordination. Put into practice, Orlikowski and Yates [100] draw from pluritemporalism in their study of a geographi- cally distributed work group and found that the use of recurrent temporal structures helps on- line communities-of-practice find a sense of symmetry—or, at least, a sense of quasi-agreement— around collective sensemaking [100]. This more expansive view of time provides a broader lens to examine the temporal affordances and constraints that continue to vex time-critical, globally distributed digital humanitarian work.

Further, pluritemporality makes visible the different modes of social time that are perceived, ex- perienced, and enacted alongside those embedded in technology and artifacts [148]. As Adam [3, p. 601] notes:

“[we] emphasize clocks and calendars or the qualitative experience of time and thus lose sight of the multiplicity of times inherent in any one social event.”

2.4 Digital Humanitarians

Digital humanitarians [84] are globally distributed online workers, generally volunteers, who spe- cialize in rapid, time-critical information monitoring, data collection, analysis, verification, satel- lite imagery analysis/mapping, and software development. These groups first came to notice as ad hoc volunteer teams in the wake of the 2010 Haiti Earthquake [131], emerging from the more formal Volunteer and Technical Communities (V&TCs) of emergency response professionals who work in collaboration with field-based humanitarian response efforts during large-scale calamities.

During the same Haiti earthquake event, a group within the OpenStreetMap (OSM) geospatial data mapping community known as HOT (Humanitarian OpenStreetMap Team) populated a nearly empty base layer map for international responders to use [125]. 25

Working as a network-within-a-network, the Digital Humanitarian Network (DHN)3 is a member of the broader V&TC consortium. With the advent of online collaborative tools, social media, and camera-enabled mobile phones, people began to form virtual networks of “everyday analysts”

[102, 132] to curate local, timely, and trustworthy information about distant emergencies to sup- port humanitarian response in the crisis zone [28, 58, 61, 76, 92, 129, 132].

Some digital humanitarian organizations, such as SBTF, work in partnership with formal insti- tutions including the United Nations Office for the Coordination of Humanitarian Affairs (UN

OCHA), NetHope,World Health Organization, M´edecins Sans Fronti`eres(MSF), and the U.S.

Federal Emergency Management Agency (FEMA), among others (The Standby Task Force, n.d.).

Digital humanitarian groups refer to their distributed online work as an activation to signal the call to action by their formal partners and to differentiate it from an on-the-ground deployment by emergency responders to the crisis zone.

In the context of distributed digital humanitarian work, time and temporality are often cited as important elements for sensemaking [102], decision support [52, 135], information sharing [29], and information quality [76]. This research in part responds to Crawford and Finn’s [34] call for critical data studies to consider the visible and invisible influences of temporality on disaster re- sponse and the ways that technologies can recursively shape crisis content. We believe that an in-depth understanding of temporal coordination in digital humanitarian work provides a new opportunity to explore CSCW in distributed teams. In particular, describing the ways that time- critical collaboration is practiced in an entirely virtual space and how new sociotemporal orders emerge to help structure qualitative representations of time. As we explore in the empirical work that follows, temporal coordination is influenced by: how time is experienced within the context of an extreme event [1, 71, 91], how time is mediated by technology [2], and how time requires significant effort to reconcile shared understanding within a globally distributed team.

3http://digitalhumanitarians.com/ 26

2.5 Methods

To explore the temporal coordination challenges present in this distributed crisis information work, we conducted a content analysis of multiple text-based artifacts constructed through the work of SBTF during Hurricane Maria. The first author has been an SBTF volunteer for four years, has conducted long-term ethnographic work with SBTF for three years, and served as a coordinator managing information collection and verification efforts during the response to Hurri- cane Maria.

2.5.1 Research Context

2017 Hurricane Maria. The U.S. territory of Puerto Rico suffered catastrophic damage after a direct hit from Hurricane Maria in the early morning hours of Sept. 20, 2017. The Category

4 superstorm with peak 135-kt winds and 9-foot storm surge made landfall less than two weeks after the Category 5 Hurricane Irma caused mass destruction on the island. Casualty counts are disputed, but the government reports that more than 1.2 million residents were impacted [30].

Seven days after landfall, FEMA activated SBTF for two Hurricane Maria-related projects: first, to gather real-time intelligence on functional medical facilities, and second, to geolocate aerial in- frastructure photos for damage assessments. The corpus of data for this study has been collected from the first activation, focused on healthcare system status. Fifty-five volunteers participated in this activation, which was conducted from September 27–29, 2017.

The Standby Task Force & Its Ecology of Information Systems. The Standby Task

Force is staffed by 2,100 trained volunteers from 106 countries worldwide. SBTF is widely re- garded for its capacity to deliver 24/7, rapid-response distributed work and information crowd- sourcing expertise to support humanitarian crisis response. SBTF volunteers collect, verify, and synthesize public data from social media, news, official communications, and satellite/UAV im- agery to produce situational awareness, remote damage assessments, and crisis maps (e.g., Figure 27

2.1) for use by emergency responders in the field [92].

SBTF, like many nonprofit organizations, appropriates assemblages of information systems to carry out their work [114, 140]. SBTF uses free, cloud-based systems in one of three contexts:

(1) as the source material of their work, e.g., as volunteers comb through websites and social me- dia feeds to identify relevant information about crisis-affected parties; (2) in the creation of new information artifacts to distribute externally, such as remote damage assessments; and (3) to co- ordinate that work, e.g., through Slack conversations.

We highlight here the most salient of these information systems, focusing, in particular, on each system’s temporal features and affordances as it relates to SBTF work practices.

Figure 2.1: A screenshot of the Hurricane Maria crisis map produced by SBTF with information collected by the volunteers and entered onto a corresponding Google Sheet.

Social Media. For SBTF, social media is a key information source. SBTF volunteers monitor social media—primarily Twitter and Facebook—through hashtags, search keywords (e.g., disas- ter name, locations), and posts by users identified as being both trustworthy and relevant (Figure

2.2). The Twitter and Facebook platforms are both based on timeline interfaces (organized by default in the user-reader time zone) which users are normed to read as real-time observations and threaded posts. While often messy, incomplete, and needing verification, social media posts by crisis-affected people and emergency responders represent the information closest to the loca- 28 tion. Thus, there is an assumption that these posts reveal the real-time nature of a disaster by eyewitnesses.

Google Drive & Google Apps Office Suite. The cloud storage platform, Google Drive, and the associated Google Apps Office Suite form the backbone of the SBTF information man- agement system. The SBTF leadership drafts various information artifacts in Google Docs, such as task checklists for standing up an activation and after-action reports summarizing each mis- sion. SBTF volunteers primarily work in Google Sheets, aggregating and organizing information in a proscribed, semi-structured template, referred to by the team as a datasheet (Figure 2.3).

The datasheet is connected to and consumed by third-party data visualization tools, such as Ar- cGIS, to produce crisis maps and other real-time data visualizations. With dozens of volunteers simultaneously entering information into the datasheet, Google Sheet’s data cell protection and custom/conditional formatting tend to fail. Therefore, on each activation, SBTF leaders cre- ate header notes on the datasheet (e.g., date of post, time, your time zone) to guide volunteers on manual entry of key temporal information, e.g., date-time formats, time zones, UTC offset lookup tables, etc.

Slack. SBTF uses the cloud-based, synchronous communication platform, Slack, to coordinate its work during activations. Slack’s display of persistent conversations (cf. [45] are typically pre- sented as chronological, single stream dialogue timestamped in the user-reader’s own time zone.

For each activation, SBTF leadership creates a variety of invitation-only channels to manage the work activities of vetted volunteers (Figure 2.4) as well as a private back-channel for the leader- ship team

GIS mapping or to provide emotional peer support by professional counselors. The Slack chan- nels are also used as a communal notification board to convey information to volunteers through links to Google Docs and other artifacts “pinned” to the details pane of each channel. The pri- mary activity takes place in the work channel. Here, volunteers dialogue with one another and 29

Figure 2.2: An example of crisis-related tweets collected and aggregated by SBTF volunteers.

Figure 2.3: A snapshot of a Google Sheet used by SBTF volunteers to collaboratively compile crisis-related information. In this Sheet, a team of volunteers is working to determine the opera- tional status of various hospital facilities in Puerto Rico. 30 the leadership team about information gathering and verification best practices and to announce their presence/departure. Archived Slack transcripts constitute an important temporal record of each activation for SBTF, as a means for producing an after-action report, evaluating its prac- tices to inform future coordination work, and as historical artifacts.

2.5.2 Corpus

We compiled a corpus of all information artifacts generated as a result of the 2017 Hurricane

Maria SBTF activation. The corpus includes the following texts: 1,549 Slack messages (107,054 words) of dialogue from the primary volunteer coordination channel transcript, 18,723 cells of in- formation from 15 Google Sheets, and one Esri ArcGIS map containing 1,407 data points. We also refer to social media data as a supplemental source relative to the Slack dialogue.

For the analysis presented here, we rely primarily on the Slack messages to understand the tem- poral discourse, as this serves as the primary communication channel for the volunteers. The

Google Sheets and ArcGIS map, as shared products of the collaboration, served primarily as ref- erents to better understand what the volunteers were talking about in the Slack conversation. In addition, we also noted the ways that these production artifacts represented time; and when and where Slack-based discourse, discussion, and disagreement around time was linked to the repre- sentations of time in these artifacts.

2.5.3 Data Analysis

We employed multiple phases of open, inductive in vivo and thematic coding and analytical memo writing [118] to interrogate the temporal work of SBTF volunteers.

We imported the primary work coordination Slack channel transcript into MaxQDA4. Two re- searchers read through the entirety of the transcripts, independently identifying all instances of temporal discourse in the data (e.g., “tomorrow”, “quickly”, “update”). The analysis team met frequently to discuss differences in coding, and to arrive at a shared understanding of what con-

4https://www.maxqda.com/ 31

Figure 2.4: A snapshot of the Slack channel used by SBTF volunteers to coordinate their dis- tributed work. Posts in the private channel include task instructions, links to activation docu- ments, and chat transcripts. 32 stituted “temporal discourse” for this domain of work. Following this initial phase, the first au- thor undertook multiple iterations of analysis, working through the data to code categories of temporal discourse; this coding was inductive but also informed by the sociology and linguistic literature on temporality [2, 68, 98, 149]. Notably, no one pre-existing temporal framework was sufficient for explaining or categorizing the diversity of temporal discourse present in our data.

As a result, we inductively derived a temporal analytic framework, inspired by and in light of previous work in this area. In our initial coding passes, we grouped similar forms of language together, wrestling with how to understand the higher-level role that the language groupings played. Given the highly contextual nature of language use, some boundaries among categories were initially quite blurred. In subsequent iterations of analysis, we interrogated our data through the guiding question: “What work is this language doing for the SBTF volunteers?”’ This guid- ing question helped to focus the boundaries of each of the six categories and helped to offer lan- guage to better characterize each discourse category. This phase of analysis also helped us move to an understanding that various forms of temporal discourse are being used as a resource for reconciling multiple temporal orders. In section 6, we present a summary of these temporal or- ders and reflect more deeply about how SBTF volunteers use discourse, often in conjunction with features of their collaborative technologies, to construct a shared sociotemporal order in which their work can take place.

We also engaged in a continuous process of reflexive bracketing [138] through extensive memo writing and discussions during the research design, data collection, and analytical phases of the study. These interpretative exercises were employed to bracket, or set aside, preconceived notions about content reflected in the Slack transcript, as a method to draw awareness to the partici- pant–researcher’s close familiarity with the Hurricane Maria activation while mitigating against preconceptions about the emerging insights. 33

2.6 Pluritemporal Analytic Framework

We found that our initial coding of the Hurricane Maria discourse from SBTF’s distributed coor- dination work did not resonate well with existing temporal frameworks developed from co-located settings. For example, we detected a particularly unique coordination burden related to the vol- unteers lack of spatial awareness with one another that was not present in other frameworks. The need to simultaneously hold several types of quantitative and qualitative time in mind, such as global locations and time zones of the collaborating team members, pace of the information flow and its relative timeliness, social media timestamps, etc., demanded a new approach.

We draw on the concept of pluritemporality [98] to give voice to a more expansive view of the many different ways that time is perceived, experienced, enacted, and embedded in time-critical, distributed crisis information work. While informed by the existing research on temporality, our inductive analysis of data in the corpus advances a broader and more holistic analytic framework of temporalities. Taken together, the six forms of temporal discourse identified in our corpus il- lustrate the complexity of distributed temporal coordination work as well as the sophistication necessary to design potential process and technology improvements in a highly abstract and en- tangled domain, like time. Here, the different forms of qualitative temporality serve to provide shared meanings at the semantic level of coordination work [120]. In the excerpts below, words or phrases coded using the complete analytical framework appear in italicized text, while words and phrases specific to the pluritemporal type under examination are also shown in boldface type, for clarity.

Time as Standard We coded language as standard time if dialogue referred to regulated, so- cially ordered formats that designate mechanized clock time. SBTF volunteers encountered sev- eral different types of time formats embedded in crisis data, including ISO-formatted timestamps in official public safety reports (e.g., 2017-09-27T15:02:19Z), social media metadata (e.g., 3:02

PM – 27 Sep 2017), culturally-specific formats (e.g., “9/27/17”, “27/9/2017”, and “2017/9/27”), 34 date-time expressions, and the informal, everyday language about time in social media posts by crisis-affected people (e.g., “around 3 p.m.”).

We noted these instances of temporal standards present in our data corpus, as well as a number of meta-discussions about these standards, including negotiations about how best to address and embody standards in the organization’s workflow:

Sabine [2017-09-27T22:48:53Z]: Okay, question that might need to be stan- dardized across deployments: When we post a date and time, no time zone is specified so we’re probably getting a jumble of different zones. Thoughts?

Time as Interval We coded language as interval if they referred to timespans, usually in re- lation to tasks, e.g., “give me a sec”, “I had a conversation . . . a few hours ago”, “morning”,

“around for awhile”, etc. Unlike formal ISO date/time standards, interval time is subjective and thus more likely to be susceptible to ambiguity and misunderstanding. Therefore, interval terms in our corpus tended to need additional explanation or elaboration as part of the workflow since they incorporated non-literal meanings and/or specific cultural contexts.

For example, SBTF volunteers extensively mined social media keywords and hashtags for in- formation about hospital operating status in the aftermath of Hurricane Maria. Posts on both

Twitter and Facebook (Figure 2.5) typically characterized the ground situation in terms of inter- val time, e.g., “a while ago” and “when”. However, these mentions often lacked key information and needed additional temporal context. From the Facebook example, “just” is relative to the poster’s time (who, in this case, is not in Puerto Rico but claims to relay information from the island). The grammatical context matters here. Using “just” as an adverb has a temporal con- notation, in contrast to its use as an adjective, which conveys judgment. Further, the Facebook user also references “2 hours” as the interval of time over which the hospital could presumably continue to operate its generator. It was not unusual to see multiple qualitative temporal repre- sentations in a single social media post that volunteers had to evaluate for meaning. 35 Figure 2.5: Interval time example from Facebook.

Time as Chronology We coded language as chronological if they referred to the linear past, present, and future perspectives of time, e.g., “already”, “now”, “today”, “new”, “later”, etc.

These terms capture an especially challenging temporal structuring problem for digital human- itarian work and information sensemaking since people use the present to base an understanding of past experience or anticipate near-future needs. This is often done to construct a timeline of events that are critical in the disaster event.

We encountered chronology words and phrases more frequently than any other pluritemporal mode. This makes sense, since the primary goal of digital humanitarian crowd work is to as- semble situational awareness of on-the-ground conditions at a given moment in time. There- fore, SBTF volunteers are constantly negotiating between collective sensemaking and the need to gather additional context to discern the state of the present situation in light of the recent past.

This dilemma is illustrated in the following exchange between Hera, an inexperienced volunteer and an SBTF leader, Rey:

Hera [2017-09-28T11:49:30Z]: @Rey I’m stressing over the details. What date do I use? The date of the textitoriginal post that stated that the hospital was open, or textitright now. Is the textittime right now, or the time from the post. What time zone do I use?

In this work, stitching together an event chronology from the cacophony of social media posts re- quires holding multiple temporalities in parallel—that is, the crisis time zone timestamps side by 36 side with one’s own relative time. Social media timestamped metadata used to curate informa- tion tends to reflect the volunteer’s own device clock, time zone, stream settings, etc. Construct- ing a chronological timeline from individual social media posts means interrogating the context behind each potential data point.

Hera [2017-09-28T11:50:29Z]: Okay Just seeing this now.

Hera then acknowledges her temporal relation to the social media post of interest as “just” and

“now” implying that the information is new and presented to her in “real time”:

Rey [2017-09-28T11:51:47Z]: No please don’t stress! Date of the original post please. Time from the post, I think as its from an entity within PR the time zone will be in AST, which is what PR follows. List the time from the post and put the time zone next to it.

The convoluted nature of assembling a chronology of events is evident in Rey’s response to Hera.

Rey, a highly experienced volunteer, also assumes the date-time of the post from “an entity within

PR” is expressed in the local time zone, Atlantic Standard Time. Some volunteers achieve a sense of pseudo-agreement with local information sources on social media by tuning their own account settings to the disaster zone time to embed themselves in local time and avoid confusion around the need to manually convert time zone metadata. Based on Rey’s advice to Hera, it ap- pears Rey may have taken this settings approach without telegraphing it to Hera.

Piecing together a chronology of events from social media data is also challenging for the volun- teers because the content and timestamps are often at odds. From this Twitter example (Figure

2.6), “yesterday” is relative since the timestamp on the tweet actually reflects the SBTF volun- teer’s Mountain Daylight Time zone and not the presumed Atlantic Standard Time of the crisis zone. However, the 10:46 PM 24 Sep 2017 timestamp, when taken at face value with the refer- ence to “yesterday,” would suggest that the author of the post is referring to the day before,

September 23rd. Once the timestamp is converted to the appropriate local Puerto Rico time,

12:45 AM 25 Sep 2017, it becomes apparent that “yesterday” actually meant September 24th. 37

The additional cognitive overhead required of SBTF volunteers to discern the differences between relative and absolute times—in context and on the fly—is quite complex and introduces troubling time- and safety-critical risks.

Figure 2.6: Chronology time example from Twitter.

Time as Tempo We coded language as marked by tempo if they convey the speed or pace at which activities take place, such as “pressing”, “quickly”, “moment by moment”, “race against time”, etc. The rapid nature of time-critical digital humanitarian work often creates a sense of urgency for the volunteers that must also be balanced against the organization’s commitment to assemble trustworthy information products for emergency responders that convey a reliable sense of ground awareness.

In this excerpt, tempo language appears more prominently when the mapping technology used to visualize the collected information interrupts the pace of Poe’s work geolocating aerial images of hospitals taken by the Civil Air Patrol (CAP). Rey responds with “wait”, “slow”, “speed” to convey her shared experience about the image upload delays:

Rey [2017-09-28T17:24:46Z]: So I put the coords as is, I ended up in Aus- tria lol. But when I limited it to two digits after the decimal point, I ended up in Castaner, which is correct. Can you try using 18.18, -66.83 ? 38

Poe [2017-09-28T17:27:26Z]: Hah! That seems to have worked - thanks @Rey! No CAP images there, unfortunately, at least on my view. :confused:

Rey [2017-09-28T17:31:14Z]: There are! Just be sure to wait a while. Ac- tually, just above the ‘Powered by Esri’ logo on the bottom right, there is the black circle that shows that the map is still uploading. Its real slow, will check if something can be done about the speed. Will keep you updated!

Poe [2017-09-28T17:31:48Z]: Oh, I see! Alright, will do! #ivelearnedsome- thingtoday

Time as Presence We coded language as presence if time is framed in spatial terms, reflect- ing how time manifests in relation to a person’s presence. This type of framing where time is rep- resented as movement within a physical axis [68] is a common approach in the study of spatial metaphoric structures in linguistics [16]. In the world, time and space are socially intertwined as an expression of location [2, 117]. Here, presence takes on a host of qualitative textures that infer distance or movement as “long time”, “catch up” and ‘‘go ahead”:

Leia [2017-09-28T06:42:20Z]: Good morning. It just took a textbflong time to read and catch up but I rebooted and hope I can move a little faster. Is there anything special I should work on or just go ahead with a hospital that has no information yet?

Time as Ephemerality We coded language as ephemeral if it acknowledged the fleeting or passing nature of time and the dynamic qualities of things happening in time. Ephemeral lan- guage often places specific emphasis on recognizing and responding to the idea that informa- tion and work practices are fluid—that everything is in motion. It also appears as an explicit ac- knowledgement that one is aware that others may have differing experiences of time, either based on their location or as mediated through the shared technologies.

Ephemeral language most often appears in dialogue about information updates or to signal the need to bring information up to date due to the rapidly changing ground truth at any given mo- 39 ment in time, irrespective of the tempo of the work. For instance, one hospital alone had 17 up- dates to its status over the activation.

Padme’s reflection, below, about chronicling and updating the ever-changing hospital status is emblematic of the volunteers’ discourse throughout the activation about the chaos in Puerto

Rico:

Padme [2017-09-27T20:45:40Z]: :sbtf: :sbtf: :sbtf: VERIFICATION is a key element for this deployment. Be sure to add the link to the original source when you add data. And double check to see if there are more updated information. In PR power is on and off, so an hospital that had power two days ago may not have it today. :sbtf: :sbtf: :sbtf:

Further, there is a texture to ephemeral time that is both relative and relational. In her post to the Slack channel, Padme concludes that the ephemerality of time complicates an already challenging task of discerning the relative currency of information (e.g., the mixed signals about whether a hospital is operating or not). Here, she alludes to the need to better align work prac- tice and technology in their effort to achieve temporal coordination. sectionTemporal Orders and Sociotemporal Disorders in Distributed Crisiswork Our pluritempo- ral analytical framework provides a thick description of the ways that distributed SBTF volun- teers use different qualitative aspects of time, in addition to standard clock-time, to coordinate their high-tempo, time-critical information work. These insights help scaffold a broader question that is the central thesis of this study: How do globally distributed teams temporally coordinate their work?

Here, we return to the concept of temporal orders introduced earlier that now come into view.

We define temporal orders as shared understandings of time within a group. People use tempo- ral orders to orient their experiences of time, which may originate in the natural world, technol- ogy or society. A natural temporal order, for example, describes the collective agreement that the lifespan of an Atlantic hurricane follows a regular pattern of formation from a rotating storm system off the coast of west Africa to potential landfall(s) to its eventual dissipation. Natural temporal orders tend to unfold as sequences in time, like a chronology, in a predictable linear 40 path with a distinct beginning and end. Alternatively, a temporal order of technology assumes a more dynamic notion of time since computers can enqueue millions of real-time data points, label them with the platform’s own interpretation of clock-time, (re-)structure the data at will, and then represent the data stream back to information consumers in any number of ways. Last, a sociotemporal order is a collective sense of time constructed by social norms and shared mean- ings that regulate activities. Put more plainly, a temporal order is a shared understanding of how different types of time manifest in our world, whereas a sociotemporal order places time into a shared social context and uses it in some form of collective practice.

Below, we interleave how the pluritemporal framework reveals the invisible work of these tem- poral orders in the Hurricane Maria activation—as well as highlighting the breakdowns, or so- ciotemporal disorders that the volunteers face in the course of their distributed coordination work. We present the results of our analysis of how work is accomplished in an environment where language about time punctuates most interactions, to understand why these different and often confusing discursive invocations of time are present in the first place.We begin by walking through an illustrative example that captures the confluence of multiple temporal discourses that shape and are shaped by temporal orders.

In the waning hours of the Hurricane Maria activation, a group is working at a fever pitch to wrap up the data collection. Two volunteers are in dialogue about how to resolve conflicting in- formation: by recency of the timestamp, credibility of the information source, or relevance to the crisis. Note that the majority of the interaction occurs within 5 minutes, and so it is rapid-fire, with many issues being discussed at once:

Ben [2017-09-28T23:41:43Z]: One of the critical things at the moment is to make sure we are only publishing one record about [each] hospital. The instruc- tions will differ based on how many people have an hour or so to work on this. Can I get anyone who can work on this to yell out here and I’ll create the in- structions.

Here, Ben, an SBTF leader located in Oceania, puts out a call to volunteers for a quick side project to help validate previously collected hospital data that appears on the crisis map. He wants to 41 ensure that the mapped information is timely and trustworthy—criteria that are sometimes at odds and require a fair amount of close evaluation by the volunteers that is difficult to achieve in the midst of a high-tempo activation. Maz, located in North America and one of the more expe- rienced volunteers, responds to Ben’s request within 30 seconds of his Slack channel post:

Maz [2017-09-28T23:42:17Z]: Well, this is the problem with putting in times- tamps that we see on our screens. The mayor made her statement at 16:51 Moun- tain. The person posting about Damas said it at 10:46 Mountain

While sorting out conflicting information is an everyday occurrence in digital humanitarian work, it is additionally complicated by the disunity between the information chronology and credibil- ity of the source. Maz responds to Ben about this tension of privileging timestamps over other evaluation criteria. She points Ben to two Facebook posts about hospital status in Ponce that were previously recorded in the Google Sheet (Figure 2.7) that are at odds. Here, there is a clash of time, as standardized timestamps and pervasive technology-enabled timelines purport to offer

‘real-time’ chronologies of information. These structures hinder the volunteers’ capacity to or- der, reorder, and reconcile a more nuanced temporal understanding of the information at hand.

Maz in part resolves this by using a timestamp from her own local time zone of which she can be assured, talking about when the mayor at the disaster site and in a different time zone made her remarks. Maz is locking down details by making references to standard time, a commonly accepted timekeeping notation. Ben replies:

Ben [2017-09-28T23:43:33Z]:We also can’t just take the timestamp as even if a post is older but is from a better source with more credible info - that is the one that should be published first.

Ben concurs with Maz—acknowledging the dilemma of reconciling different information. This discursive exchange gives way to a shared understanding of how taking the chronology at face value is entangled with the need for a more sophisticated evaluation and verification of the con- tent posted to Facebook. However, this does not merely entail a hierarchical ranking of the value 42 of certain metadata over other descriptive attributes but a reckoning with the complexity of the information, the precision required to assess it, and how to parse it on the fly.

Maz [2017-09-28T23:44:15Z]: The later record is the mayor. I have no idea who the other person is and how she would know about every hospital in Ponce

Within seconds, Maz replies to Ben, sharing her similarly aligned view: A Ponce elected official is a more trustworthy informant than a local Facebook user who posted a more recent but conflict- ing status report that lacked detail about the information’s provenance. Thus, source credibility is assessed to supersede the social media timestamp as an evaluative judgment.

Ben [2017-09-28T23:44:33Z]: We will have two layers in the map - one show- ing the most relevant info only and one showing all. So everything will be used but FEMA will just be able to use what they want easier.

Ben [2017-09-28T23:44:50Z]: It’s not just about this one example.

Having agreed about two dimensions of the information—recency of the timestamp metadata and credibility of the source—Ben introduces a new wrinkle: the relevance of the information to the unfolding crisis. Here, he suggests generating map layers as a way to temporally arrange the information in ways that allows consumers of the information (in this case, FEMA) to decide what to prioritize: timeliness or dated-but-certified accuracy.

Maz [2017-09-28T23:46:33Z]: I understand that but how do you propose find- ing the correct one within the time constraint? I found the ”right” one by re- reading both the posts.

Maz pushes back on Ben’s leap in adding relevancy as a new dimension for reaching agreement about the hospital status information. While they agree on the contextual precision needed to integrate clock time and credibility into their common agreement, she reminds him about the tempo of the work, which until this point in their dialogue was unacknowledged as a real tension and threat to the larger goal of producing timely and trustworthy information. She continues: 43

Maz [2017-09-28T23:58:53Z]: The only way I can see these kinds of discrep- ancies to check these is to sort the data. I can copy the sheet into a separate workbook and sort by hospital name. But I don’t think there’s time, if we only had an hour to begin with. I have to be back on the fire in an hour. Let me know what you would like me to do

The SBTF Hurricane Maria project officially ends shortly after the discussion between Ben and

Maz. However, a new problem is revealed about the Ponce hospital status discrepancy during the data verification process, where a small team of trusted volunteers spot check information en- tered onto the sheet at the conclusion of the activation. Upon further investigation, the SBTF volunteers discover that there are more than four hospitals in Ponce (in contradiction to the mistranslated statement attributed to the mayor in the sheet) and that the “other person” who posted on Facebook at 10:46 Mountain Time that two hospitals were closed is actually a Univi- sion reporter based in Puerto Rico using her personal social media account to amplify her news reporting. So, in this case, both reports are correct and both sources are credible, but strictly re- lying on timestamps causes a misleading interpretation. To paraphrase Reddy et al. [112, p. 50],

While time is certainly important, context also matters.

The tempo of digital humanitarian work varies considerably, adding yet another layer of complex- ity to distributed coordination work for ad hoc teams. Each activation establishes its own unique pace according to the complexity and urgency of the tasks, access to information, and, crucially, the availability of volunteers to coordinate with one another. Even with a globally distributed team, the nature of this work is still subject to the everyday work–life rhythms of the volunteers.

As Maz notes, Ben’s impromptu and quick-turnaround data validation side project runs head- long into the realities of a much more complicated, time-delimited task. She offers a data sorting workaround to speed up the task as a way to reconcile a temporal disruption in the work with her planned departure to attend to other work obligations.

In this rapid exchange covering a number of issues in digital humanitarian work, we see three primary temporal orders to which the pair attends: the temporal order of the crisis with respect to the damage it has done to hospital infras- 44 tructures and the constantly shifting state of operations caused by new natural hazards, e.g.,

flooding, mudslides, etc.; the temporal order of the technologies, which mediate the source information of their work—in this case, the timestamp metadata of social media posts about hos- pital status; the sociotemporal order of the distributed work that influences the urgent pace of the coordination and the limited window of time that Maz can work on the task.

The remainder of this paper will discuss how SBTF is constantly undertaking temporal coordi- nation work to establish some degree of workable and shared temporal agreement. But that work begins from a state of temporal disorder by virtue of the arrangement of a globally distributed work team searching for information about a natural hazard event that runs on its own time and information that is mediated through technology platforms that associate their own temporal metadata with it.

Figure 2.7: Conflicting hospital status reports.

2.6.1 Three Temporal Orders

We find three primary temporal orders present in distributed crisis work: (1) the temporal or- der of the crisis; (2) the combination of temporal orders of the various technologies being used; and (3) the sociotemporal order of the distributed team. These temporal orders vary from one an- other due to differences in the composition of the social groups implicated in each, as well as the 45 nature of their relations to other temporal orders and markers of natural time. All three tempo- ral orders manifest in the discourse that is central to distributed crisis work, particularly through the work required to orient team members to align the orders, and do so across the team. By revealing each of these largely tacit temporal orders, we are able to more specifically articulate and target those temporal features that provide the clearest opportunities to help resolve friction points and improve globally distributed team work practice and technology adaptation.

Temporal Order of the Crisis. The most foundational temporal order to the digital humanitar- ian work context is that of the crisis itself. Here, the shared understanding of crisis time within the group is dominated by timekeeping and the passage of time in a linear progression of crisis phases. For instance, the crisis event is located in a particular time zone—in this case, Atlantic

Standard Time (AST). This includes communications about the hurricane landfall and other hazard-related events, the rescue phase activities, the recovery phase activities, and so on. Days and times in news reports and social media posts are often communicated with an implicit as- sumption of having taken place in the AST time zone. All of this necessitates frequent conver- sions between “crisis time” and the time zones in which the digital humanitarian work is being done. Further, temporal references in news and social media are often either implicit (e.g., “Hos- pital Mel´endezBayamon will run out of diesel ...”; Figure 2.8) or lack context that would facili- tate precise organization or interpretation.

This problem increases volunteers’ reliance on the timestamps of social media posts. However, media timestamps are typically displayed to reflect the time zone of the user-reader, not that of the user-poster who originated or shared the post (Figure 2.9). Although this design decision can facilitate rapid assessment of the recency or staleness of a post, it can also result in confusion. It is quite common for linguistic time references in the social media content to not align with the displayed time stamp.

Natural disasters can additionally be divided into three broad, sequential phases: pre-impact, impact, and post-impact. Neal suggests that each phase is characterized by “a continuum of dis- aster,” or period of time when the usual everyday rhythms of life are disrupted by the immediate 46

Figure 2.8: Mentions of time in crisis-related social media posts are often implicit and/or non- specific.

Figure 2.9: Temporal metadata differences due to user-reader locations. 47 priorities imposed by the disaster event [91]. Variations in individuals’ experience of this facet of crisis time manifest in the content, tenor, and informational focus of their social media commu- nications; characteristics that, then, affect the work that the distributed humanitarian team can accomplish.

Temporal Order(s) of the Technologies. In distributed humanitarian work, cloud-based tools and platforms are used to generate and synthesize information and metadata into temporal artifacts, such as datasheets, coordination transcripts, and crisis maps. Each of these tools imposes its own temporal order that is understood by SBTF volunteers as dynamic and malleable due to filter- ing, sorting, and layering, as well as algorithmic interventions. For example, when social media platforms control when posts appear (and in what order), they create their own sense of tem- poral organization. Information is made available based on inaccessible algorithms, internally maintained clocks, and database schemas, all of which may introduce changes to the informa- tion’s temporal metadata, divorcing the data from its original state. Additionally, by changing the ways that time is displayed to users, systems can either distort or clarify temporal relation- ships. In some cases, technologies can be configured to foster a sense of common ground among people [26] through shared representations of time. However, in most cases, these technologies are implemented according to commercial priorities.

Furthermore, the illusion of “real-time” is a source of significant tension in temporal coordina- tion in distributed crisis work. These tensions manifest in two primary ways: 1) the false sense of immediacy embedded in technology-mediated information and 2) the limited ways that time is represented in social and collaborative computing technologies. For example, synthesizing social media as a reliable, actionable, and timely information source for real-time situational awareness has proven to be quite challenging [126]. Social media platform infrastructures are not especially conducive to (nor founded upon an economic model that encourages) displaying an ordered real- time stream from the deluge of posts from and about the crisis zone.

Another key temporal disruption is the inability to represent different modalities of time in off- the-shelf, cloud-based collaborative technologies used to collect and synthesize qualitative crisis 48 information. Although SBTF has, in the past, experimented with using customized collaboration support tools, the overhead of distributing, training, and provisioning long-term technical sup- port for maintaining these tools is high (see also Voida et al, [140]).

As we noted in our pluritemporal framework, people employ multiple forms of time references side-by-side to construct a sense of order, as illustrated in this text message about the deteriorat- ing conditions of a hospital posted to Twitter (Figure 2.10). Information about hospital status frequently inferred qualitative features of time, such as urgency, priority, and updates to previous information. Yet, data structuring tools, like Google Sheets, lack the extensibility to accommo- date complex temporal sensemaking strategies, aside from storing different types of standardized timestamp formats. Instead, this crucial information gets lost in text fields, further degrading a shared understanding of “crisis time” by the SBTF volunteers.

Figure 2.10: Qualitative representations of time in social media.

Sociotemporal Order of the Distributed Work. Volunteers contributing to a globally distributed digital humanitarian endeavor are, by definition, spread out across immense geographical space.

As a result, they are also spread across time, joining into the collaboration from a variety of time 49 zones, day and night, during breaks in daytime employment and in the midst of discretionary evening time. Not only do volunteers have to reconcile their own time to the duration of and milestones imposed by an SBTF activation, they also have to construct shared understandings of how to coordinate with one another to accomplish that work. The rapid micro-coordination

[75] reflected in the exchange between Ben and Maz, above, illustrates how SBTF volunteers— one located in Oceania and the other in North America—are able to order and negotiate work within their shared, computer-mediated environment.

2.6.2 Temporal Coordination: The Work of Reconciling Sociotemporal Disorder

Breakdowns in finding adequate points of reference across the three temporal orders are the cause of much temporal discourse that we found and coded in our corpus. Instead of achieving an “or- der” in which they can organize their work from a shared understanding of time—their own so- ciotemporal order—they are instead working from a state of “disorder.” SBTF volunteers must work discursively to reconcile temporal orders, both for themselves and collaboratively with oth- ers in their distributed work team to reach agreement about what time it is where.

The tyranny of clock time imposes deadlines, shift schedules, and production expectations on work. For broadly dispersed global teams, coordination must be negotiated across day/night rhythms, biological clocks, time zones, and localized sociotemporal orders, like periods desig- nated for work and leisure. For 24/7 organizations like SBTF, which does not structure its work in shifts or formal units of time, this creates an additional layer of complexity added to tempo- ral coordination. Volunteers come and go at will; therefore, the dynamic nature of what consti- tutes “work time” is sociotemporally disordered from the start. Erratic work rhythms and con- stantly evolving team formations hinder individuals’ capacity to develop an awareness of what co-workers are doing and when.

The following exchange between SBTF volunteers illustrates sociotemporally disordered dis- tributed teamwork. Early in the activation, the volunteers were provided inconsistent instructions about how to note on the Google Sheet which hospital, from a list supplied by FEMA, that they 50 chose to search. The volunteers either did not see the instructions pinned in the Slack channel or were confused by the directions and forged ahead using previous practice that turned out not to apply in this case. At this point, seven days have passed since the hurricane first made landfall in

Puerto Rico before FEMA activated SBTF. Hospitals and other major social institutions on the island remain in a state of chaos, which is uncharacteristic at the one-week mark of the disaster impact phase. Moreover, the information request by FEMA required a new, unfamiliar workflow for SBTF volunteers to collect highly targeted, real-time information about hospital status with- out duplications. Here, the disrupted temporal order of the crisis is out of sync with the typical rhythms of the sociotemporal order of the distributed work:

Sabine [2017-09-27T23:11:10Z]: Guys, please sort this out, because the in- structions on the sheet aren’t clear, and there are initials columns in both the “list of hospitals” and the “social media sources” sheets.

Sabine [2017-09-27T23:12:41Z]: I’m trying to point myself in a focused direc- tion and hoping others know to do the same in whatever way is most coherent. :slightly-smiling-face: To be completely clear: Nothing is more frustrating to me than wasting time doing something someone else is already doing.

Here, Sabine complains about the lack of coordination among the team by specifically calling out

“time wasting” as a threat to team productivity and shared understanding (“doing something someone else is already doing”). Absent a fix to the temporal disorder, she follows up a few min- utes later with a more urgent call to action for the community to clarify the parameters of the immediate, time-critical task, re-iterating time as a precious commodity:

Sabine [2017-09-27T23:17:01Z]: Clues might be in the version history, @Rey, but for the moment I would just like some clarity so I can proceed in a non-time- wasting fashion. Pretty please, y’all. :slightly-smiling-face: Because we need ev- erybody on the same page in order to not waste time, and that means everybody should be assigning themselves a hospital (or however many) from the list itself, if I’m understanding correctly.

Sabine’s frustration boils over, which instigates a new strategy—peering back into the Google

Sheet version history to resolve the confusion about the current, improvised work practice that 51 deviates from previous hurricane activation instructions. Sabine suggests to Rey, an SBTF leader, to review the content timeline from the instruction-sharing tab in the Google Sheet to determine when the breakdown occurred. Here, the temporal order of the technologies is invoked to reveal the friction between the temporal order of the crisis and sociotemporal order of the distributed work that is driving the dispute.

Lexi [2017-09-27T23:17:07Z]: I’m afraid I’m off, this is a short but tough one that has had to be be put together very quickly, with technical issues thrown in for extra fun! Have patience and work together! Love you all :heart-eyes: I will be back for the last few hours tomorrow :starspinning:

Lexi enters the conversation (though has been working steadily in the background) to offer miss- ing temporal context about the behind-the-scenes tempo of the activation launch that likely caused the disruption. Note that she is also signalling the end of her own personal work shift, with some regret about the implications for collectively completing this task (“I’m afraid I’m off ”). In urg- ing the volunteers to be patient, Lexi draws attention to the need to slow their pace as a poten- tial remedy to the unfolding temporal conflicts between the crisis and the work.

Rey [2017-09-27T23:19:27Z]: Please go ahead and check the history while I check with the [activation coordinators] to get some finality on this matter.

Mace, one of the project coordinators, interjects herself into the continuing disagreement by us- ing presence time references (“come from”, “come up”) to appeal for a new, agreed-upon work-

flow to leave the disordered temporal crisis, technology, and work rhythms behind.

Mace [2017-09-27T23:26:46Z]: Hi Sabine, I hear where you’re coming from but the hospital list tab is feeding a map. Too many people on that tab risks possible data corruption. How do you think we can come up with a way to stake claims on search terms without interfering with that tab?

2.7 Discussion: Designing to Support Temporal Coordination Work

In this analysis, we highlighted how SBTF volunteers employ conversation to work through and around sociotemporal disordering and to establish some degree of sociotemporal agreement so 52 their work can productively unfold. Much like other forms of articulation work (e.g., Schmidt &

Bannon [119]), the negotiation about time and establishment of a shared understanding of “what time is it?” is carried out as an essential part of work practice but operates invisibly in the mar- gins to support the functioning of the broader mission and aims of the organization. The diver- sity and scope of the discourse that transpires to support temporal coordination is striking. The derivation of the pluritermporal framework presented in section 5 arose from our analysis of all the ways that people communicated the different modes of time that are at work in this problem space as they are perceived, experienced, and mediated via technology—often problematically.

We use these same pluritemporal categories to scaffold a set of design implications that could reduce the amount of discursive articulation work needed to overcome sociotemporal disorder, especially in time- and safety-critical events that are increasingly mediated by technology.

Designing for Time as Standard, Time as Interval, and Time as Chronology:

Tools for automating timestamp management and revealing data/metadata

The prevalence of conversations around standards, the resolution of ambiguous or underspeci-

fied temporal intervals, and the reconstruction of a coherent chronology, highlights the dearth of useful information about temporality provided by current collaborative systems: timestamps— represented in varying base time zones—or vague characterizations of recency. Most temporal information with which the volunteers we studied were working is conveyed descriptively, embed- ded in the social media content itself. An accurate ordering of the social media content and the events that they describe has to be reconstructed manually—and has to be negotiated, aligned, and ordered discursively among the volunteers. The implicit assumption embodied in these— admittedly mundane—facets of interface and infrastructure design is that [re-]presenting tempo- rality with seemingly objective, but experientially indifferent timestamps, is sufficient.

Timestamps, however objective and reliable as they may appear, create problems of their own.

Although SBTF coordinators currently attempt to pre-structure data collection sheets like the 53 one appearing in Figure 2.3, the brittle nature of Google Sheets’ automated formatting rules for dates and times regularly leads to breakdowns in the face of dozens of volunteers simulta- neously editing the document. In the discourse under study, we observed, for example, com- ments about the interjection of different localized date representations when a row was added to a shared Google sheet, debates about the lack of uniformity in 12/24-hour time entries, and regular disagreement about whether timestamps should be conveyed in the local crisis time zone or a “universal” (e.g., UTC) time zone, given the lack of explicit conversion tools to and/or from the crisis-local time zone.

These problems also raised questions about data trustworthiness, necessitating manual workarounds, such as continuous quality-control checks of the sheet by designated, experienced SBTF volun- teers. These standards-oriented and temporal interpretation-oriented problems might be ad- dressed by enhancing the time and date handling capabilities of tools like Slack and Google Sheets.

Having the ability to globally specify—and lock down, for all but SBTF activation leadership— an expected timestamp format and timezone offset for a sheet or a conversation, and then pro- viding the capability to automatically convert (or suggest conversion) to these shared norms would be an asset to large numbers of contributors working together to make sense of temporal sequences. The ability to quickly reveal relative time information (e.g., conversion of a time to

UTC or to display a timestamp offset from one’s local time zone or from a reference datum such as the time of crisis onset or the scheduled conclusion of an SBTF activation) would serve as use- ful informational resources for volunteers. This would reduce many of the questions about time zones and timestamp formatting that we observed in the Slack transcripts.

Another way that the collaborative computing platforms used by SBTF volunteers could better support adherence to standards, resolution of ambiguous intervals, and reconstruction of chronolo- gies would be to provide better mechanisms for capturing and representing the provenance of the data upon which this work is based: when Facebook posts or Tweets are referenced, for example, the temporal and provenancial metadata surrounding those posts could be rendered visible, per- haps upon request, and it could be included as data are referenced, moved, and/or aggregated. 54

What would it mean, for example, to be able to cut and paste a tweet into a spreadsheet and have the spreadsheet formulas and/or macros in that sheet have access to the temporal history of the post, as well as the ability to perform computation (e.g., conversion to standard time rep- resentations and a primary time zone, validation or visualization of the recency of content, etc.) on that content on the user’s behalf? What would it mean for “data detector”-like technologies

[90] to be able to not only detect but also provide suggested resolutions for ambiguous mentions of time based on the known temporalities in which that information was constructed, propagated through a social media network, and then retrieved? What if this kind of rich temporal informa- tion were accessible to technology end-users instead of being relegated to esoteric JSON represen- tations of the postings, intelligible only to sophisticated platform API users?

This kind of nuanced temporal information—ideally, pre-sorted chronologically by the platforms, themselves—might be embodied within specialized crisis-response “portals” that allow for more direct, comprehensive, and non-timeshifted access to social media streams in partnership with the major social media companies, themselves. Alternative approaches that would also address some of the shortcomings that we observed in our analysis include development of crisis-focused social media aggregation tools, in the vein of consumer-oriented apps like HootSuite5 or Tweet-

Deck6, or the creation of browser plugins to allow for on-demand interrogation of the information provenance and in-depth metadata records associated with individual tweets accessed through a typical social media website.

Designing for Time as Tempo: Tools for promoting task and activity awareness

Technologies or interaction techniques to complement or obviate the need for volunteers’ ob- served discourse about and around the tempo of the work (both relative to and in the context of the crisis and the SBTF activation) are more challenging to design for. Interfaces like Time

Aura [78] suggest how real-time pacing indications can be ambiently added to a primary task in-

5https://hootsuite.com 6https://tweetdeck.twitter.com 55 terface and provide peripheral cues about how the pace of work is (or is not) matching expected norms or paced to meet an approaching milestone or deadline. However, these interfaces can also increase individuals’ sense of time pressure and anxiety by rendering small deviations from tem- poral expectations or norms quite visible. Effective design solutions in this domain need to strike a subtle balance between providing the kinds of informative cues that are currently missing from shared tools like Slack. Likewise, doing so in a way that enables volunteers to support one an- other in settling into a sustainable pace that does not compromise the integrity of the shared output or the health and well-being of the volunteers themselves.

Implicitly, conversations about tempo and pacing also reflect a lack of strong shared awareness about how finer-grained activities and tasks fit into the larger picture of the crisis response acti- vation. Feeling “rushed” or expressing concern about how the most important analysis activities will all be accomplished—at an acceptably high level of quality—within the 72-hour window that

SBTF targets for completion of their intensive digital humanitarian response suggests that indi- vidual volunteers don’t have a particularly clear sense of how or where their contributions fit into the bigger picture. Activity awareness tools and/or collaborative task-management dashboards

(e.g., Muller et al, [87]) integrated into Slack or overlaid on the coordination-oriented communi- cation channel might also help to address and mitigate these instances of uncertainty and anxi- ety.

Designing for Time as Presence and Time as Ephemerality: Tools for supporting sociotemporal and collaboration awareness

Temporal coordination work that is made visible through language that expresses features of presenceand ephemerality in order to reconcile sociotemporal disorder—signals the presence of similar work coordination challenges that we see in CSCW work about remote co-presence and awareness [40], with the added complexity of supporting a relatively large-scale team for a rela- tively short duration of time. Here, systems that provide increased situational and task aware- 56 ness would help to off-load much of the temporal coordination work that we observed. Although the technologies employed by SBTF to coordinate and compile information—Slack and Google

Sheets—do include representations of who is online, these representations provide little context besides indications of synchronous presence.

Techniques like media spaces [11] could implicitly provide better cues about activity and tempo- ral constraints (e.g., time of day, location in which work is taking place), these approaches are generally not intended to scale beyond a few close collaborators. Other indicators of engagement, activity, and activity history in persistent conversational spaces have also been proposed (e.g.,

Erickson & Kellogg, [45]). Adding similar visualizations—for example, the amount of time a vol- unteer has been active, their anticipated amount of time remaining available to participate in the current working session, a coarse representation of their current activity such as a link to the most recent contribution on the shared Google Sheet or the hashtags of the last social media post examined, and, as a counterpart, how recently a particular spreadsheet cell had been “updated” and by whom—might help to create a more tangible sense of “presence”, cooperation, and coor- dination among volunteers.

Overall, exploring design implications through the lens of temporal discourse categories also sug- gests an initial design space for supporting temporal coordination work among globally distributed teams. Here, our pluritemporal framework helped us to identify several dimensions of automation and awareness support that SBTF’s current toolset does not support well: automated timestamp management, data and metadata awareness, task and activity awareness, and sociotemporal and collaboration awareness. The emphasis here on supporting distributed work through different kinds of data, task, and collaboration awareness resonates well with prior CSCW research on de- sign strategies for productively supporting group work without imposing formalized workflows.

Additionally, the multiplicity of facets of awareness that we see emerging here suggests that fa- cilitating other kinds of awareness might solve additional challenges beyond those that we uncov- ered by focusing specifically on temporal discourse. A more comprehensive design space analysis of the role of different kinds of awareness tools might be a valuable next step in designing to sup- 57 port these kinds of globally distributed collaborative teams.

2.8 Conclusion

Returning to the introductory excerpt in this article, we see once again how Leia is working to construct a legible chronology of a complex and dynamic crisis situation based on snippets of information originating in a distant part of the world, a region that is operating within its own temporal disorder in the wake of a disaster:

Leia [2017-09-28T09:46:06Z]: I just made sure all the hospitals the [Depart- ment of Health] says are up and running as of yesterday had at least 1 report that they were open. It certainly is confusing as there are social media reports seemingly from the same time saying they are not open.

The pieces of information that Leia has found have been re-interpreted through the combined technological temporal order of the social media platforms that delivered them, adding layers of ambiguity about when the information was posted and making it difficult to interpret the appar- ent conflicts in hospital operational status. Leia uses language about time in a plurality of ways to articulate these tensions. She also uses language about time to signal where she is in the midst of completing this task, inviting an opportunity to reconcile temporal disagreement among the globally dispersed volunteers monitoring the shared chat channel. Language like this illustrates a key challenge of distributed humanitarian work and highlights opportunities for directly address- ing the breakdowns introduced by these competing temporal orders through design solutions.

This research has sought to understand how globally distributed teams temporally coordinate to accomplish their work by studying team discourse in high-tempo digital humanitarian informa- tion work. Our empirical analysis began with the development of a new pluritemporal analytic framework derived from the linguistic analysis. Subsequent analytical passes at a more macro level of analysis resolved how the discursive references were associated with the three different temporal orders that the team was invisibly wrestling with—the temporal order of the crisis, the combined temporal orders imposed by the technologies employed by the team, and the so- 58 ciotemporal order that the team itself tried to construct in its own virtual workplace environ- ment. Members of the Standby Task Force discursively worked toward shared temporal under- standing to “order” the sociotemporal disorder (not once, but continuously throughout the acti- vation) so as to achieve moments of temporal coordination. From these results, we, in turn, inter- rogated the technical origins that are in part attributable to these sociotemporal disorders.

The study of temporality in distributed work has taken on increased empirical and theoretical at- tention within the CSCW community—and with good reason. Temporality is an essential feature of technology, work practice, and human experience. And, as we have shown in this research, temporality and the alignment of temporal orders represents a significant challenge for individ- uals who are simultaneously distributed across time and space. Our work, bridging between the empirical and theoretical, also lays the groundwork for a more intentional shift towards design- ing for temporality within the CSCW community. Our contributions of a pluritemporal analytic framework, an articulation of the temporal orders implicated in this kind of distributed humani- tarian work, and an initial discussion of design space axes that might serve as scaffolding for fu- ture design and implementation work all point towards opportunities for the community to con- tinue to engage with these vital challenges. Chapter 3

Narrating Time as Temporal Sensemaking

3.1 Introduction

Advances in information and communication technologies (ICTs) have spurred people to come together online to collaborate, socialize, and share. Over the past decade, ICTs have also played a key role in the emergence of digital humanitarians—an altruistic, global volun- teer network that provides information support to emergency responders during large-scale humanitarian crises caused by natural disasters, disease epidemics, and the mass displacement of people [84]. Online volunteers gather, curate, verify, and analyze public information about the crisis zone through this specialized form of crowdsourcing. These groups rely almost exclusively on the appropriation of free, cloud-based ICT tools, e.g., collaboration platforms, GIS mapping software, and social media sites, to meet their situated needs [133].

Until recently, these homebrew systems worked well enough[140]. However, with the frequency and intensity of natural hazards on the rise and an historic 71 million people forcibly displaced from their homes due to disasters, conflict, persecution, and climate change [139], the scope and scale of high-tempo, time-critical distributed information group work is becoming much more challenging to coordinate across time and space.

Beyond crisis informatics, time is a salient concern in the broader domains of human-computer interaction (HCI), as well as social computing. A sociotemporal order is a framework that struc- tures and ascribes shared meanings about human experiences in which we put time into prac- tice [149]. Sociotemporal orders are depicted by calendars, timestamps, time zones, chronological 60 social media streams, etc., then enable distributed work practices across cloud-based, collabo- rative information and communication technologies (ICTs) [73]. The quantitative and qualita- tive aspects of time each provide helpful insights into the form and function of ICTs. Yet, there has been little direct attention in the literature on how distributed teams reflect on, make sense of, and reconcile as a collective narrative these different types of time that exist in tandem in computer-supported coordination work (CSCW).

One digital humanitarian volunteer described the ICT dilemma this way:

“People talk in stories, but response teams talk in spreadsheets, charts, and maps. In order to extract actionable information from average citizens, the community- generated data has to be collected and processed in a standardized, structured, and organized way.”1

In this research, we investigate an extreme example of temporal breakdown in distributed infor- mation group work: the visible and invisible ways that digital humanitarian volunteers construct a “real-time” narrative of the 2017 Hurricane Maria impact in Puerto Rico. These insights are drawn from semi-structured interviews that reveal a process of triage, evaluation, negotiation, and synchronization to enact temporal sensemaking in high-tempo, globally distributed, time- and safety-critical information crowdwork. Finally, we present design implications for cloud- based, collaborative ICTs to better support collective construction of temporal narratives as sensemaking devices in digital humanitarian work. This research extends prior work on tempo- ral coordination in this domain [93] by allowing digital humanitarian volunteers to describe in their own words how they, individually and as a team, enact temporal sensemaking.

3.2 Background: Digital humanitarian work

“Everyday analysts” across the world first mobilized at scale in reaction to the catastrophic 2010

Haiti earthquake [102]. Online volunteers formed ad hoc groups to collaboratively produce geospa- tial mapping, SMS shortcode analysis, and Creole language translation within hours of the mas-

1Quotations are presented verbatim with minor corrections for clarity and mid-quote shortening represented by [...]. 61 sive 7.0-magnitude temblor [84, 125]. In the intervening 10 years, the field of crisis informatics, and particularly, these highly-networked digital humanitarian groups have developed deep ex- pertise in crowdsourcing, crisis mapping, and information support [6, 28, 58, 61, 92, 106, 126,

129, 132]. For this paper, we turn our focus to The Standby Task Force (SBTF)2, a 24/7 rapid response information support group with 2,100 volunteers from 106 countries. SBTF has been called into service by the United Nations Office for the Coordination of Humanitarian Affairs

(UN OCHA), U.S. Federal Emergency Management Agency (FEMA), and other international

NGOs. SBTF appropriates free, cloud-based, and ubiquitous ICTs to conduct its information work, like many nonprofit organizations [114, 140]. As a data source, SBTF manually collects and curates public information from social media platforms and indexed websites for relevant news about the crisis. Google Drive, Google Apps Office Suite and ArcGIS serve as the informa- tion management and artifact production system. The group additionally uses Slack, a cloud- based synchronous communication platform to coordinate their work. The first author has ex- tensive experience with SBTF in two capacities as a: 1) volunteer coordinator of 15 activations3, including Hurricane Maria, and 2) researcher conducting long-term ethnographic studies with the group since 2016.

3.2.1 Case: 2017 Hurricane Maria

Homes, businesses, and public works infrastructures on the Commonwealth of Puerto Rico were destroyed after a direct hit by Hurricane Maria on Sept. 20, 2017. The category 4 superstorm made landfall on the small U.S. territory, just two weeks after Hurricane Irma had earlier caused widespread flooding and utility outages across the island. SBTF partnered with FEMA to col- lect, assess, and map public data on the real-time operational status of hospitals and health- care facilities across Puerto Rico. The 72-hour rapid response activation was conducted between

2https://www.standbytaskforce.org 3SBTF prefers the term activation to distinguish its virtual information support from an on-the-ground emer- gency response deployment. 62

September 27-29, 2017. Within 36 hours, SBTF reported reliable information on 33 high-priority hospitals. By the end of the activation, SBTF produced for FEMA verified status reports on 63 hospitals, or 93% of those on the island, as well as 44 other critical healthcare facilities, including nursing homes, dialysis clinics, blood banks, and morgues.

3.3 Related Work

As Thomas et al note: “...the ways in which people conceive of and experience time are often at odds with the ways in which interactive systems represent and express temporal factors”[136, p.

3303]. The human side of experience-representation disconnect can be broken down into three component parts: social constructions of time, distributed coordination, and temporal sensemak- ing.

3.3.1 Social Constructions of Time

Clocks, timestamps, time zones, live streams, and time-related artifacts, like schedules, enable people to share a common understanding about time [51, 150]. With these artifacts, time still must be calibrated and then agreed to. Clocks embedded in mobile phones, on the other hand, provide both precision as well as a single source of time which influences quite differently how we socially align activities [74]. These analog and digital social constructions of time allow people to establish rhythms, patterns, routines, and other types of ordered behavior necessary for coordina- tion.

3.3.2 Temporal & Distributed Coordination

Besides having a common agreement about telling the time, people also need to be able to act on their shared temporal knowledge in order to synchronize in temporal coordination across the team [44, 82, 112]. For distributed teams of remote, non-colocated and/or global workers, es- tablishing a common framework for clock-time is quite challenging over time zones and diurnal phases, let alone creating organizational and communication structures to achieve distributed 63 temporal coordination with one another over time and space [100]. An especially relevant con- cept in this context is Crowston and Kammerer’s study of the “collective mind”, a perspective that describes how individuals develop an understanding about their own contribution to group work [35]. Extending prior theoretical work by Weick and Roberts, collective mind is established through a process of new member socialization, opportunities for conversation, and recapitula- tion of organizational stories that are repeated, reinterpreted, and reinforced for newcomers and veterans alike [146].

3.3.3 Temporal Sensemaking

Temporal sensemaking is a complex cognitive and social process that integrates common agree- ment about a particular social construction of time, e.g., past, present, and future, with shared meaning to create a retrospective sense of order through collective interpretations and explana- tions about ambiguous information [145, 146].

Dawson and Sykes [38] suggest that sensemaking is inextricably linked to temporality due to “a backward glance for retrospective explanation” and one “that seek[s] to reconstruct events from the past.” However, during an unexpected crisis, time is not experienced as an entirely linear phenomenon. Waller and Uitdewilligen [143, p. 188] argue that in chaotic emergency scenarios,

“sensemaking is of crucial importance for reducing confusion, guiding action, and preventing or- ganizational disintegration.” However, crisis information is revealed over time and not completely at the onset. The reality of information flow curtails the group’s sensemaking and capacity to distribute their knowledge.

3.3.4 Temporal Narrative

Narrative storytelling as a sensemaking device is temporal. Boje [15, pg. 106] describes orga- nizations as “collective storytelling systems in which the performance of stories is a key part of members’ sensemaking process and a means to allow them to supplant individual memories with institutional memory.” Cunliffe et al [37] define a temporal narrative as a series of collectively ne- 64 gotiated interpretations of an event over multiple points of view, multiple contexts, and multiple moments in time. Here, Zerubavel’s concept of sociotemporal order [151] provides some clarity about how temporal narratives are organized and the role they play in coordination work. So- ciotemporal order serves as a collective framework for establishing agreement about time to en- able productive work and to put time into practice. This sense of order and agreement is also crucial for structuring a coherent narrative within a shared social context. However, there is an important point of contestation in the temporal narrative literature about the role of sensemak- ing that this study examines. Here, previous theoretical and empirical work [37, 36] question the utility of Weick’s classic description of sensemaking as a linear, retrospective rationalization of what just happened. Rather, Cunliffe and Coupland argue that a temporal narrative must also take into account real-time experiences. From this perspective, temporal sensemaking and the stories that flow from these events are threads interwoven in time—responsive to the future and the present as well as the past.

3.4 Methods

Highly abstract concepts, like temporal coordination, are quite challenging for people to discuss.

Therefore, we used in-depth semi-structured interviews to understand the social and cognitive intersections between temporality and crowdwork among the distributed team members.

We designed the interview protocol to flow alongside archival documents from the Hurricane

Maria activation, which were posted to a password-protected website and accessible only by the participants. We posted verbatim chat excerpts from the SBTF Slack channel and a copy of the SBTF Google Sheet used for data entry; both of which included several examples of com- mon temporal sensemaking problems. Grounding the interviews referentially in archival but nev- ertheless familiar documents helped to overcome the high level of abstraction about temporal sensemaking by prompting participants to have a dialogue with the document excerpts about specific time-related work practices and technologies [41]. This grounding-in-excerpts approach also helped to re-immerse participants in the details of the digital humanitarian activation that 65 occurred two years’ previously, serving to bring past interactions into focus.

3.4.1 Participants & Recruitment

We used SBTF’s Hurricane Maria hospital status activation roster to initially screen 55 volun- teers who signed up to help. While people are often eager to help during an SBTF activation, like other types of volunteer endeavors squeezed in between work, family, and other obligations, there can be discrepancies between those who initially register and those who are able to make the time to actively participate. We reconciled these particular challenges by focusing participant recruitment on a subset of volunteers who actively contributed data to the project.

Based on this screening criteria, we targeted our outreach to 29 volunteers who contributed at least 3 entries to the hospital status data collection sheet. This subset of volunteers devoted sub- stantial time to the 3-day project (Mdn = 8 hours; range: 1-32 hours), according to self-reports at the conclusion of the activation. Further, this volunteer pool also represented a mix of experi- enced hands, as well as those working on their very first activation.

Nine SBTF volunteers responded to interview requests. Participants were not compensated for their time. We continuously reviewed the interview data for novel insights and stopped recruiting participants once we reached saturation, e.g., repeated descriptions of work practices that yielded no new information.

3.4.2 Data Collection

Each interview began with a grand tour question [128] about the volunteer’s overall Hurricane

Maria activation experience followed by a mini tour question about their information search ap- proaches to discover the post-hurricane functional status of hospitals in Puerto Rico. Subsequent questions explored different information work practices during the hurricane activation, includ- ing how each volunteer: discerned time-related data/metadata, assessed the recency, timeliness, and relevance of information, reconciled discrepancies and conflicts in the data, and managed time/tasks. We prompted participants to offer perspectives about their own individual work 66 practices, those of the group, and the cloud-based, collaborative tools used to collect and record the hospital data. The interviews concluded with a “magic wand” question to elicit concerns and recommendations for improvements to work practices and technologies.

The first author conducted all of the semi-structured interviews over Skype, Slack or telephone, per the participants’ preference. The median interview length was 108 minutes (SD = 33.3, range:

33-143 minutes). Each interview was audio recorded and transcribed for analysis. The interviews yielded 14.5 hours of in-depth discussions with SBTF leadership and volunteers from Europe,

North America, South America, and South Asia.

3.4.3 Data Analysis

We used a constant comparison method to inductively identify recurring concepts from each in- terview transcript [24]. Themes emerged through simultaneous close reading, coding, memoran- dum writing, sorting, and comparing each transcript for similarities and differences. Drawing from Lincoln and Guba [72], we worked to establish trustworthiness through several methods.

First, the first author is highly engaged in SBTF, as both a volunteer and primary investigator of a four-year ethnographic study of the organization. Next, we constructed interview questions to elicit commentary by relaying different points of view from other participants, enabling them to reflect on the triangulation of their own experiences, discussions of group values and practices, and reflections on excerpted chat transcripts and collected data from the activation. Last, the

first author used a reflexive bracketing process [5, 138] of memo writing and discussions with the research team to mitigate against preconceptions about the emerging insights.

3.5 FINDINGS

Distinct patterns of temporal sensemaking practices emerged through our thematic analysis.

Here, we present four approaches to temporal sensemaking—triaging, evaluating, negotiating, synchronizing—that were prevalent in the descriptions of high-tempo, time- and safety-critical information crowd work. These sensemaking approaches are often used in combination to con- 67 struct a temporal narrative from the disparate information collected by the SBTF volunteers as a way to convey shared understandings about health care access on the ground in Puerto Rico.

These new insights extend prior CSCW research about how people talk about and navigate time at work and in their everyday lives [44, 73, 100].

3.5.1 Triaging Temporal Information

Information triage is “the process of sorting through (the possibly numerous) relevant materi- als, and organizing them to meet the needs of the task at hand. The term ‘triage’ seems partic- ularly appropriate for this activity, since it is often time-constrained, and requires quick assess- ment based on insufficient knowledge” [79, p. 124].

All of the volunteers spoke at length about engaging in various triaging processes to order and prioritize the firehose of information from social media and news reporting that was generated in the aftermath of the hurricane. Several volunteers specifically used the term “triage” to talk about their initial sensemaking work during the early, chaotic moments of the crisis response.

P8: You have to triage a lot of sources to really get an idea of what could be happening.

During temporal information triage, the volunteers described several influences—some that are in their control and others that are not—on their ability to prioritize and order large, cascading amounts of information from diverse sources. Each of these factors influence the ultimate value of the chronological narrative they are attempting to construct for responders.

For instance, the hurricane trajectory creates a temporal flow of data that impacts volunteers’ ability to triage time-sensitive information.

P1: I’ve noticed is that there is kind of an information life cycle dur- ing these emergencies ... There is the moment when the hurricane hits and that moment is a sort of blackout. You do not have much information coming because you do not know what to expect. And then after that, the hurricane has passed, you have this moment again where there’s no new information coming. It may last a few hours or several days. It depends on the impact that the hur- ricane had because if there is no power, you might wait several days before you 68

start to get some information. And then you start to have some sort of data flow here and there and then out of the blue you start to have an overload of informa- tion ... The first hours there’s nothing, there’s nothing, there’s nothing then out of the blue, there’s everything.

The inevitable information overflow demands a particular sensibility to the triaging process— acknowledging when the information search no longer serves the primary goal of crafting an ac- tionable narrative for responders. This sense of when to stop chasing dead ends, let go, and move on to mine other information sources generates a fair amount of stress for SBTF volunteers due to the high-tempo pace of the triaging work and their strong desire to help vulnerable people while also contributing to a trustworthy report.

P5: One of the other difficult things that we would run into is triage, of course. And there’s probably multiple senses of that term for just this sort of effort, but at what point do you say we can’t verify this particular piece? Is there an easier one to work on that we may be able to deliver accurate work faster? Which sucks because ... it kind of feels like abandonment. Where could I point myself where I can actually get a result that’s usable that maybe helps somebody sooner?

Crisis information triage is often referred to as “finding a needle in a haystack” [39, 76]. Prior work in this domain has typically focused on the development of automated search and filtering tools and less on peer-to-peer crowdsourcing best practices for teams in which the overhead for developing and maintaining technical tools is high (c.f. [140]. The SBTF volunteers, however, voiced a need for more low-fidelity sensemaking support for crowdsourcing work informed by ad- vanced technology.

P1: I think that one thing that we need to remember and that I really would like to change is to remember that technology is not the solution for every- thing. This is something that we always struggle a bit with our activators because they say, okay, you have kind of machine learning and then the machine will do everything.

P5: There’s got to be some good templates for just how to get familiar with what’s on the Internet, what’s useful, what’s credible. How you can start making those 69

on the fly judgment calls—which is what we were doing every second these de- ployments. Honing your skills requires having, let’s say, a good algo- rithm in your head that’s properly calibrated to assess things. It’s on the fly analysis. Where do you go to get the info and then how you analyze it ef- fectively?

3.5.2 Evaluating Temporal Information

Once different hospital status materials are triaged, the volunteers set about to assess different temporal aspects of its informational value to contribute to the developing narrative. Despite the globally distributed nature of their work, each volunteer followed remarkably similar temporal sensemaking evaluation practices. All of the volunteers begin with a rapid assessment of social media or news article timestamp to determine its recency. If the information was deemed “fresh,” then more scrutiny was placed on exploring other temporal cues and gauging the reliability of the source beyond the initial triaging process.

However, simply relying on timestamps was often a mixed bag for temporal sensemaking. The volunteers frequently mentioned the variation in timestamp formats across different online mate- rial, the nebulous information represented on updated news stories or edited social media posts, and the lack of timestamps on certain types of source material, like institutional press releases or hospital website announcements. Some used clever workarounds by comparing timestamps with other temporal references in the information.

P6: I would definitely be looking at timestamps and trying to figure it out. I dunno, I guess trying to make those judgments about the credibility of the people, whether they were there, whether it was firsthand information. Or whether it was something that they read or heard. If it’s a news article, it can be from much earlier. So I don’t know. I looked for ones that had specifics like where they said, ’I tried to go over there at 2:30 and [the hospital] was closed’ or something like that.

FEMA’s request for real-time hospital status was often at odds with the representation of time displayed on the timestamp, which created an additional sensemaking burden.

P5: First of all, you’ve got to make sure you understand what the times- tamps mean on the sources. For us, it was so important to nail down as 70

close to real time as we could. But information gets stale so you can’t always nec- essarily tell.

Unexpectedly, the volunteers talked extensively about using different work-arounds to triangulate temporal and non- temporal information to enhance the sensemaking process, especially with respect to the credibility and reliability of when something was reported to have happened.

P6: I would look for time related details in the text of the tweet, if they actually say what time it was in addition to just the timestamp on the tweet. But if they said, ’this morning as of’, if they actually mentioned a time, then I would find that much more reliable than seeing one where they gave details about an event but no time in the text. And if that was only relying on the tweet times- tamp of the tweet. That would be less reliable to me for time.

P8: When you search for information this way, you triangulate many things at the same time. If I found a tweet, what was this person tweeting before? Up to now? Is it a real person? Is it a bot? What does this person do? What are they interested in? What do they know about, right? How fresh is this information? Because in normal life, maybe a tweet two days ago could be very fresh for certain things, but working right now on this hurricane, a tweet from two days ago, is old. It’s probably almost useless.

Additionally, the challenge in using triangulation as an additional and more complex evaluative sensemaking layer is that triangulating is also beholden to the same rapid tempo of data collec- tion as the triaging process with a much heavier cognitive load to construct a real-time account from the information at hand.

P8: I have two contradicting things. I can find more information so I can tell you what really happened. But this is what I have. Maybe you find something else and with both actions, then we can go into some place with some certainty. But, right now, I don’t have it.

3.5.3 Negotiating Breakdowns in Temporal Information

SBTF volunteers encountered numerous breakdowns that impacted their capacity to make sense of time-related information from large volumes of confusing material to triage and evaluate, to unreliable timestamps that frustrate their attempts to construct real-time narratives. Not only 71 do these breakdowns need to be brokered on the fly by the volunteers but they need to be dis- tinguished between currently actionable information versus potentially useful future information that may reveal trends or predictive outliers based on the hurricane impact lifecycle. Or as P5 remarked, We get a big bucket of tea leaves to read!’

Social media is a powerful source of information about individual stories of resilience and suffer- ing, as well as the state of infrastructure and social conditions during a crisis. People are often compelled to communicate and share information as a way to feel helpful or to process trauma

[104]. However, the lack of internet access and/or electric service typical during a catastrophe contributed to lags in online reports, sometimes in quite lengthy periods of time, about the func- tional status of the hospitals. SBTF volunteers also frequently observed unaffected family and friends publicly posting on social media private text messages from those in the crisis zone, which also caused real-time report delays. As a result, volunteers expressed skepticism about whether timestamps on social media posts accurately reflected current ground conditions.

P3: So when somebody posts something it may not correlate to when they experienced it. And you know, this is part of the problem with what we do.

Temporal breakdowns can also be quite complex, involving multiple pieces of conflicting data, timestamps, and equally reliable sources. Every volunteer recounted frequent instances of con-

flicting hospital status reports due to the chaotic situation on the ground. One negotiating tactic a volunteer described was to re-engage with the source materials derived from the initial triage, evaluation, and triangulation processes but re-contextualized against the current backdrop of the hurricane response trajectory, as time frame noted by a different volunteer in the triaging section above.

Here, the volunteer describes how they searched for the most current hospital status by anticipat- ing information breakdowns;areas where power, internet, and cellular service were still out would be less likely to post real-time accounts compared to sources in locations with proximity to tem- porarily restored services. 72

P2: And also, internet usage in an area where, you know, it’s going to hit to see whether we would get information or we wouldn’t get information from there. So if there’s an area where the internet usage or social media usage is less than that then you know most of the information is going to be posted to government sites or news portals and all that. And then you look at those sources of information. You see if some relative from another place is posting information about them, so it does make a big difference in the kind of information that you are looking out for. You get that sense of where exactly you need to look.

And, yet, unresolved temporal breakdowns can also be helpful for sensemaking purposes in chal- lenging situations:

P2: You know, sometimes even to have conflicting information is infor- mation.

3.5.4 Synchronizing Temporal Information

Once an information breakdown was identified, the volunteers described various ways that they worked to resolve the issue and to synchronize their sensemaking process with other volunteers to promote shared work practices and values.

Volunteers would frequently seek out advice from more experienced hands in the Slack channel on how to reconcile contradictory information as a way to resolve their own sensemaking prob- lems as well as a strategy to build SBTF common knowledge.

P4: ... part of it is to work with the volunteers to have them understand that both can be true. And that’s part of the problem they don’t have any sort of big picture way to see it.

Later, this long-time volunteer described how they shared their own sensemaking process to guide another volunteer struggling with a complex temporal breakdown. The negotiating process often involved a loop of triaging, evaluating, and triangulating additional information followed by an- other round of re-negotiating to either resolve the conflict or verify the existence of parallel infor- mation.

P4: Now, let’s see if we can figure out a story for this one. Does that give you enough intuition that you can do some more on your own that’s not seat of the pants. Really digging down ... [because] they are both possibly true. 73

Constantly engaging and re-engaging with material was a familiar strategy shared by the SBTF team. Here, an experienced volunteer described a rapid-fire cyclical approach to reconciling in- formation as well as advice for “newbie” volunteers on how to develop their own sense of intu- ition about the validity of social media posts to add to the collective narrative. The process of

“tracing backward” to locate the original post was a common tactic among the SBTF volun- teers. To them, these repetitive activities helped sync up the current information to the “origin story” from the shared post, which then evoked greater confidence in their findings by producing a chronological through-line to link the more recent information.

P4: You also find times when things that are retweeted or shared from Face- book that can be quite old. So, I’ll also try and trace back to the original post or whatever it is to get a sense of when that is. Then, also look at it chronologically, in the context of when and where the person who posted it pushed it out. It’s not a linear thing. You have a hospital, you find something, you chase it down. Doesn’t really work, doesn’t make sense, doesn’t tell me enough. Try again. Find a different source. Oh, that makes more sense. And that informa- tion is duplicated over here. Now, I go back to the original one. Does that make more sense? It’s so circular, iterative, and intuitive.

There is a strong collective culture within the SBTF corps. All of the volunteers spoke at length about the camaraderie they experienced with the team and the importance of learning from one another through collaborative workflows, shared work practices, and collective knowledge. The ability to synchronize their work across a large, globally distributed team was attributed to the openness of the community, the ease of using Slack to communicate, and the core values of being nice to one another—even when that may complicate the information collection.

P4: I think because [the data] is so complex, checking in with other people on a regular basis with the complicated ones really adds a whole other way of thinking about it. That’s really crucial. So talking to other people.

P4: When [data entry errors are] egregious, we’re very good during [the activa- tion]. I think we’re good at doing it in a humane way where we’re not pointing somebody out. Somebody has a private conversation and you do it kindly. So, I think, for instance, we do a good job of making sure that people aren’t breaking confidentiality or privacy issues. ’You know, if you did this, it would really 74

make a difference. It will be a little bit harder, but look at how much better the data you would be collecting would look.’ But it’s a hard thing to do in any circumstance.

Across all the interviews, volunteers raised concerns about the ephemerality of the information in relation to the objective to produce actionable intelligence to on-the-ground emergency respon- ders. Several volunteers challenged the notion of “real time” as an unrealistic expectation of the activating partner. Pushing back further, one volunteer expressed frustration with the value and usefulness of a temporal narrative with a brief duration.

P9: So this is all problematic with the hospital status in Puerto Rico is that I have a sense that this, the value of that information we collected had a very short life.

3.5.5 Crafting the Emergent Temporal Narrative

Cunliffe describes a temporal narrative as a series of collectively negotiated interpretations of an event over multiple points of view, multiple contexts, and multiple moments in time [37].

In this context, the temporal narrative is constructed from synthesized information on the fly, of unknown provenance, in chaotic and uncertain conditions, and produced by a globally distributed team far from the crisis zone. In this way, the temporal narrative constructed by the volunteers is a synthesis from multiple points of view, contexts, and moments in time. The temporal narra- tive then serves as a function of SBTF’s collective sensemaking and a product for sensemaking by the activation partner over time.

P6: You get a sense of the overall, bigger picture in that area without just look- ing at one post at a time. So kind of a way ofsynthesizing the collective in- formation that’s being posted about that particular point.

The rapid assemblage of an actionable temporal narrative about the hospitals highlights the tight integration necessary between the crowdsourcing process (triage, evaluation, negotiation, and synchronization) and the collaboration tools used (Slack, Google Sheets, ArcGIS, and social me- dia platforms) in order to produce timely and trustworthy information. 75

P9: It’s meant to be something that can be done very rapidly before the others understand what is happening. Puerto Rico is an example of a deep dive into analyzing what these messages are really about.

P5: We’re working to a deadline. I think with a lot of this, we in our position are always going to be stuck with good enough and playing the odds and mak- ing judgment calls, hoping that it’s right.

P2: So then if anything was not credible, then, what I would have done was till I actually find another source of information that can confirm that I would choose not to publish it, to not give any false information, potentially false in- formation there.

Another key aspect of the emerging temporal narrative is crafting it to address the information sensemaking needs of the activation partner. Here, the volunteers reflect on recalibrating their temporal sensemaking approaches in order to not lose sight of the complexity and messiness of the data as a mirror of the chaotic ground situation.

P4: I think keeping in mind the needs of the user, is something that we do have to keep remembering that it’s not whether or not we get caught up in a particular tragedy or particular story, but whether if we put ourselves in the shoes of the user of the data, will they want to know that?

P4: Very often I’ll say something to [volunteer name] that it doesn’t really seem like it’s adding much. I’m not quite sure which one to do. And [volunteer name] will take one look and say ‘you’d make a different decision knowing both of those things than you would knowing only one of them. So both should be there.’ He really brings it back to the user at the other side. Being able to provide the kind of information the user at the other side is going to need and sometimes they need complexity.

P2: I mean at the end of the day, despite all of these sheets, [the responders] ended up looking at the map and seeing what kind of action to take.

3.6 Speculations on new temporal sensemaking approaches

The work of SBTF is mediated (and often constrained) by the appropriated assemblages of free, off-the-shelf cloud-based collaboration tools they make use of. The lack of extensibility in repre- senting a wider range of temporal factors or inability to better accommodate their information crowdwork practices is a source of frustration for the team. 76

We concluded the interview sessions with a “magic wand” question as a cool-down exercise for the volunteers where they were asked to creatively speculate about improving SBTF work prac- tices and/or appropriated technologies. We present a small sample of their informal design ideas indexed to the temporal sensemaking practices—triage, evaluation, negotiation, and synchroniza- tion. These brainstormed ideas point to a set of sociotechnical design implications to speed up the construction of timely and trustworthy temporal narratives for future activations.

3.6.1 Triage Tools to Scale Good Will

The manual nature of the information triaging process is both a source of frustration and point of pride for the volunteers. Most volunteers prefer to stay closely connected to the flow of infor- mation, as overwhelming as it might be at times, to help them stay current on hospital condi- tions, anticipate trends, and/or spot meaningful outliers. Only one volunteer imagined a semi- automated, human-in-the-middle system of collecting social media posts via API through an elaborate system of dynamic keyword filtering. Other volunteers offered contrasting micro- and macro- visions to increase the power and speed of human computation.

Micro: Avoid being too prescriptive

P5: I can’t predict how easy it is to implement because it depends on volunteer interest and availability. It is what we were talking about with educating people about the kinds of resources to look for. Any kinds of analysis to do. They’re al- ready kind of getting it but a little bit more, I don’t want to say standardized, be- cause that’s not exactly what I mean. ‘Cause, you know, sometimes the standard becomes a pair of blinders.

Macro: Scale up the organization

P8: I get the feeling that the way to scale is to make an organization that could be more distributed. . . being able to scale up somehow making semi-autonomous cells that amplify our work. In order to do that, each volunteer becomes a little coordinator within a network. What kind of organization do we need if we really like to scale this up: 10X? 100X? How can we tap all this good will, time, abili- ties, and availability of devices and connections? What kind of tools do we need to make this incredibly distributed powerful thing work? 77

3.6.2 Visual Representations of Time & Evaluative Metadata

Various visual representations or ambient displays of ratings, confidence levels and dynamic tem- poral heat maps that convey additional evaluative data about the hospital reports entered on the

Google Sheet was suggested by several volunteers. The new metadata embedded in these displays could provide interesting new forms of human- and/or machine-computed qualitative evaluation, analysis, and insights. Some of the ideas are akin to those advanced in Temporal Design [108] and the slow technology design agenda [99] to radically rethink social coordination.

Relevance score or confidence level filtering

P6: Maybe adding a field that is you know, where you score the post for rele- vance or something so then you could sort on those.

P5: We could confidently verify [by] classing info by our confidence level.

Temporal heat map + slider combo feature

P9: So you use a filter to filter out everything ... and then you use the time slider to see how dots on the map become strong or weak. They fade out when it’s over, they fade in when your slider is approaching a high activity.

3.6.3 Negotiate Clustered Information Rather than One-offs

Many of the thornier information breakdowns faced by the SBTF volunteers become even more intractable due to the relatively flat data structures presented in Google Sheets and social media platform interfaces. Lacking the ability to reassemble the data without interfering with the work of other volunteers in the Google Sheet inspired several volunteers to think about how to cluster data to more easily negotiate stale or missing data. subsubsectionGamify the Google Sheet data entry to transform it into a relational database

P9: As a volunteer, you could work very fast without bothering too much about the structure. You should relate to only your data to stitch them together. And then behind the work you are making it relational database. 78

Trend visualization over time and place

P6: I’m just thinking about that right now. I was thinking because like if you have multiple reports coming in at different times. Showing spatio-temporal data on a map is hard. But it might be interesting to use clustering or something like that where you, when you click on the cluster, it opens up all the different reports and it could highlight all of the records that are associated with that point. So for a single point at that hospital, it may have been entered five different times be- cause you found five different reports for that hospital. So if you clicked on the point, it would then promote to the top of the table, the five records that pertained to it.

3.6.4 Synchronization Across the Team through Feedback

A desire for more feedback was mentioned during all of the interviews. In particular, volunteers were interested in knowing on a higher order how their work contributed to the collective mind of the activation. Likewise, they wanted to imagine solutions to find a better balance for vol- unteers who are not able to participate for long periods of time but would benefit from timely micro-activation updates to feel helpful and be productive in assembling the group temporal nar- rative.

Periodic activation digest

P6: Often volunteers come on for an hour or something and they’re not going to read through the last 23 hours of Slack posts to get up to speed. So just kind of putting out little public service announcements on Slack every hour or something so that everybody’s seeing that information, you know, kind of a little digest of all the important things that we learned over the last hour.

Collective mind + operations dashboard

P6: ... setting up dashboards that are kind of tracking everybody’s statistics, kind of gamifying it a little bit and showing a leaderboard of contributions. People seem to really like that and respond to it. And not just the, like, who’s ahead, who’s posting the most, ‘cause we don’t want people just going crazy just to get their numbers up and they’re not being very careful. But also just the overall group accomplishment. Like, how many we’ve collected, you know, the numbers of posts and . . . show the progress getting toward that goal. So sometimes those 79

kind of graphic things are just motivating ‘cause you can see that you’re making progress and that you all collectively making a dent in this thing. But, also, there was some healthy competition.

3.7 Discussion and Implications

In our analysis, we identified four approaches to temporal sensemaking—triage, negotiation, eval- uation, and synchronization—that SBTF volunteers use to collectively construct a temporal nar- rative as a sensemaking device to organize time- and safety-critical information during the Hurri- cane Maria humanitarian crisis in Puerto Rico. In this case, the construction of a temporal nar- rative is a practice-centered approach that unites social interaction, articulation work, sociotech- nical structures, and collective sensemaking to enable productive, high-tempo distributed work that quickly evolves over a short period of time [35, 119].

Taking a more granular approach to understand more precisely how SBTF volunteers make sense of and construct a collective understanding from high-tempo crisis information revealed interest- ing synergies between Weick and Roberts’ [145] formative work on retrospective sensemaking and the present-oriented narrative framework espoused by [36]. The globally distributed digital hu- manitarian work conducted by SBTF interleaves the collective wisdom of prior activations, the immediate situational awareness in the crisis zone, as well as anticipating the trajectory of in- formation common to certain natural hazards, in this case a hurricane response. Several of the interview excerpts noted in the previous section illustrate how sensemaking is also influenced by the rapid cycling of the past, present, and future of the crisis itself—a temporal influence that the SBTF volunteers must reckon with in order to make sense of the information produced by the crisis.

In this context, the frameworks posited by Weick et al. and Cunliffe et al. become less a matter of competing descriptions and more of an opportunity to view sensemaking as a continuum to understand the rhythms of crisis information that rapidly unfold over time. For instance, as one volunteer lamented: 80

P9: The value of that information we collected had a very short life.

The pace at which sensemaking must take place for SBTF volunteers blurs the demarcations be- tween retrospective sensemaking and a narrative constructed in the flow of the moment, as Cun- liffe and Coupland suggest, when the lookback may be measured in mere hours.

A speculative design idea that emerged from the interview with the same volunteer as above, illustrates how classic sensemaking and collective narrative research can work together to con- struct timely and trustworthy crisis information over time.

P9: So you use a filter to filter out everything ... and then you use the time slider to see how dots on the map become strong or weak. They fade out when it’s over, they fade in when your slider is approaching a high activity.

Here, this data visualization tool would allow volunteers to amplify their data story by control- ling access to retrospective information contained in the crisis map. The technology, in turn, also becomes its own resource for temporal sensemaking. The filter mechanism of this hypothetical tool provides another level of fidelity for volunteers to triage the data, evaluate the metadata, negotiate breakdowns and inconsistencies, and then synchronize a common story about the infor- mation between the team. Thus, bringing together classic understandings of sensemaking with a more “tangible” collective narrative that can aid rapid knowledge-building during crisis response.

3.8 Conclusion and Future Work

Time is of the essence in humanitarian crisis response to speed help and ease suffering. However, making sense of time embedded in crisis data in extreme high-tempo, time- and safety-critical scenarios is not well-supported in work practice or ICT tools. Using the backdrop of the 2017

Hurricane Maria in Puerto Rico, we interviewed digital humanitarian volunteers to better under- stand how these globally distributed teams construct a coherent temporal narrative about condi- tions on-the-ground. Our analysis contributes four approaches to temporal sensemaking, includ- ing the triage, evaluation, negotiation, and synchronization of dynamic and often contradictory 81 information toward the construction of a temporal narrative as an information support artifact for responders. This research extends the canonical organizational studies work of Karl Weick and colleagues on sensemaking with communication studies research that seeks to explore how interpretive processes, like narratives, enhance how people respond to and make meaning of life

[36]. Future work further exploring and prototyping the volunteers’ speculative design ideas are planned to facilitate improvements in distributed work practices and cloud-based collaboration technologies to help produce timely and trustworthy crisis information products. Chapter 4

Designing for Sociotemporal Disorder

CHAPTER 4: Designing for Sociotemporal Disorder

4.1 Introduction

Sociotemporal disorder occurs when there is a lack of collective agreement about the social orga- nization of time [93]. These breakdowns in common meaning about “what time is it?” and how to put this shared knowledge into practice can range from minor annoyances while scheduling a meeting across multiple time zones to potentially catastrophic synchronization errors in critical systems.

Discordant temporal experiences can happen for a number of reasons, including a lack of discur- sive precision about time that contributes to miscommunication about activity coordination, vari- ances in timekeeping standards, dynamic social situations where time is fluid and ephemeral, etc.

[96, 95, 94, 93, 97]. This misalignment between human experience and instrumented time is a particularly intractable challenge for sociotechnical systems to cope with. Reconciling sociotem- poral disorder toward a collective understanding that enables productive work is a surprisingly complicated endeavor for which there is little guidance from prior research in human-computer interaction (HCI) or computer-supported cooperative work and social computing (CSCW).

In this chapter, we explore an extreme case of temporal coordination that frequently involves dif- ferent forms of sociotemporal disorder in order to push the boundaries of these entanglements.

Here, we focus on globally dispersed, high tempo, time- and safety-critical digital humanitarian 83 information work [28, 58, 61, 84, 92]. This work provides an interesting testbed for theorizing about sociotemporal disorder because it must wrestle across domains: multiple temporal orders that occur in the crisis itself, the improvisational nature of the distributed work coordination, and the everyday technologies appropriated to do the work [93].

Take, for example, the algorithmic shuffling of a social media timeline—a crucial information source for digital humanitarian work. Here, posts are presented out of temporal sequence and embedded with altered metadata dissociated from its original state. Instead of displaying posts as a linear timeline in reverse chronological order, the default presentation of “top posts” is typ- ically used to advance the commercial interests of the platform by promoting messages predicted to generate high levels of user engagement. Consequently, digital humanitarians must impro- vise to establish a sense of time—that is trustworthy yet dynamic enough to reflect the chaotic ground conditions in the crisis zone (see Chapters 2/3).

The sociotemporal disorder of these curated posts is further disrupted by the embedded tempo- ral metadata that reflects the new algorithmic order, now dissociated from its original state. As

Rader and Gray [110] note, while users are often aware of the temporal mismatch in the orga- nization of their timeline feed, the perception remains that algorithmically-curated social media content still constitutes a “real-time” chronology of activity occurring with their social networks.

This discontinuity of machine-filtered content, modified metadata, relative user clock-time, and user perceptions introduces a multi-pronged sense of sociotemporal disorder.

Sociotemporal disorder can also be amplified by a lack of fidelity in sociotechnical systems to rep- resent time other than as uniform, instrumented, and quantifiable. Multi- dimensional qualitative manifestations of time, such as collective rhythms, events, and symbolic temporalities, are preva- lent in work, life, and society [2, 56, 71, 88, 98, 127, 151]. Yet, these rich, varied, meaningful, and ubiquitous temporal experiences are still inadequately supported by information systems Erick- son2016, Lindley2015, Mazmanian2015, Norris2019, Wajcman2008.

This research builds understanding about: 1) How might globally distributed teams reconcile the sociotechnical breakdowns entangled between distributed temporal coordination work and appro- 84 priated collaborative tools? and 2) In what ways might collaborative technologies better support the temporal coordination work of globally distributed teams? The contributions of this research are four-fold:

• A novel virtual participatory design (vPD) protocol to engage distributed, online-only

teams in the process of sociotechnical system co-design;

• A descriptive account of the different forms of sociotemporal disorder that our digital

humanitarian informants encounter and must reconcile;

• An analytical framework to interrogate sociotemporal disorder in distributed work prac-

tices and the appropriated sociotechnical tools that enable their work; and

• Practical implications to guide the design of cloud-based sociotechnical tools to propel

the future of remote work in useful and more humane ways.

While several studies point to sociotemporality as a crucial factor in digital humanitarian work

(e.g., [28, 34, 132], this topic has received little direct attention in the literature. Further, this research also provides additional grist for understanding the larger and more complex practice of technology appropriation in CSCW [133]. Drawing on my prior research described in Chapters

2–3, this final study in the dissertation is designed to help address these gaps.

4.2 Digital humanitarians as a community of practice

Digital humanitarians are networks of globally-distributed online volunteers who collect, curate, and produce data artifacts to support on-the-ground emergency response during a humanitar- ian crisis. Here, we again focus our attention on the digital humanitarian team, the Standby

Task Force (SBTF)1, as our site of study. SBTF has partnered with leading crisis response or- ganizations, including the United Nations Office for the Coordination of Humanitarian Affairs,

1https://www.standbytaskforce.org 85 the U.S. Federal Emergency Management Agency, the U.S. Coast Guard, Doctors Without Bor- ders/M´edicinsSans Fronti`eres,etc. The community is composed of 2,100 volunteers from 106 countries worldwide. Due to its broad global dispersion, SBTF‘s hallmark is its unique ability to provide round-the-clock 24/7 assistance to its partners.

When SBTF is called into action, we refer to the project as an “activation” rather than a “de- ployment” to differentiate that the work is conducted online and not in the field. SBTF coor- dinates all of its work through the asynchronous chat platform Slack and assorted cloud-based sociotechnical tools, e.g., Google Office Suite, social media platforms, etc.. Akin to other crowd- sourcing and peer production groups, SBTF maintains a fairly flat organizational hierarchy. An elected Board of Directors officially governs the U.S.-incorporated nonprofit organization but typ- ically does not participate in activations. The Core Team consists of highly experienced volun- teers who interface with partners, determine the scope of activation requests, and manage the activation. Coordinators are also volunteers with significant previous activation experience. They assist the Core Team in the management of the activation and engage directly with volunteers but typically do not interface with partners. Volunteers are the backbone of the activation work and take their direction from Core Team and Coordinators. Some volunteers are quite experi- enced, while others may be participating in their first activation after completing SBTF’s requi- site screening, orientation, and information-knowledge work training.

SBTF is also an example of an online community of practice [69] because the volunteers share a collective identity, learn from one another through interactions, and have a repertoire of collec- tive knowledge, language, and resources that allows their work to productively unfold [147]. This distinction matters in how SBTF structures its learning-based practices because HCI design ap- proaches must be able to: 1) engage online-only communities in the design of sociotechnical sys- tems that they have come to rely on to conduct their work and 2) appreciate the learning process itself as a deeply held value in the community. 86

4.3 Applying participatory design to sociotemporal disorder

As an HCI approach, participatory design (PD) is grounded in an interactive, learning-by-doing approach [13, 19, 66] as is action research [62], design science [59], and other classic HCI design methods that are especially attentive to practice, problem-solving, and generative artifacts in in- formation systems. For our purposes, PD was a natural fit, in large part, due to its emphasis on ethnography as a means to reflect community values, understand the work environment, and pro- mote user empowerment. With five years of ethnographic study on SBTF already underway, the design component of this study could easily bridge back to the previous work on the theoretical descriptions of sociotemporal order/disorder and temporal sensemaking, and the pluritemporal analytical framework to disambiguate temporal meanings in talk.

Participatory design (PD) is a practice that was popularized in Scandinavian HCI research in the

1980s as a collaborative approach to help workers and unions influence technological changes at work—initially in the skilled trades where automated systems were first introduced and later in office environments with word processing and software-enabled technologies [12]. Its pragmatic roots in 1970s cooperativism offered a radical new perspective of field-based, action-oriented de- sign that engaged the notion that communities of practice had expertise and value to offer the design process. [14, 17].

A primary advantage of the PD approach as a human-centered design method is its emphasis on user empowerment and engagement [27, 87]. More recently, however, Bødker and Kyng [14], early pioneers of PD, have become quite critical of the current state of PD research. They ar- gue that PD has become too focused on collaboration for collaboration’s sake, to the exclusion of confronting the power dynamics inherent in the development and implementation of technolo- gies in the workplace and society. Further, the process appears to be less focused on learning and empowering participants as a key outcome and, in their view, has become overly committed to methodological purity to the exclusion of working to evoke long-term change and/or advance am- bitious new technologies. 87

This critique is also echoed by Saad-Sulonen et al. [116] who challenge researchers to consider future-oriented and processual temporal lenses of PD projects over time. In particular, they ar- gue there is a need to consider a continuum of phases of participation and not simply launch PD studies as one-off participatory endeavors. This perspective is especially salient to my work with

SBTF and roles as a participant, coordinator, and researcher committed to digital humanitarian work and its evolution as a community of practice. Blomberg and Karasti [9] draw a similar dis- tinction in ethnographic research about computer-supported cooperative work (CSCW) practices that is also quite relevant to the PD study:

We believe that multi-sited ethnographies not only allow for studying practices of technology production and use that are distributed in time and space, but also for design that stretches out over temporal and spatial horizons (e.g. Karasti et al., 2010). In this we encourage more ‘integrative’ ways of connecting ethnography and design. [9, p. 406]

In response to this call for emergent new design practices, this study addresses that methodologi- cal gap to develop new ways to engage with online communities of practice.

4.4 Methods

The challenges of constantly reconciling sociotemporal disorder in the midst of a crisis activation generates friction for SBTF volunteers despite numerous attempts at different sociotechnical ap- propriations, makeshift approaches, and manual workarounds in their work practices. The team has long acknowledged that these ad hoc approaches are less than optimal and have expressed openness to considering new ideas that meet two conditions: 1) reduce the cognitive stress on the volunteers in data collection and sensemaking and 2) advance the production of high quality, trustworthy information for activation partners.

4.4.1 Overview of the virtual PD approach

To address these issues, we developed a novel virtual participatory design (vPD) protocol specif- ically adapted for a globally distributed remote team as a way to provoke productive collabora- 88 tion, thoughtful identification of problems/obstacles, and creative solutions across a wide range of SBTF volunteer demographics and prior activation experiences. The vPD protocol was con- ducted with 17 global SBTF volunteers from Europe, North America, Oceania, and Southeast

Asia who were assigned to 5 small teams according to their proximate time zones. Both quan- titative and qualitative data was generated by each team during the synchronous vPD work- shops, including individual team Slack chat transcripts [141, 152], online whiteboard sketches, post-workshop feedback surveys, screencast video recordings of the chat and whiteboard activi- ties, and field notes. The data was analyzed using a variety of interpretative methods including affinity diagramming, prototype design sketching [60], field note reflections [47], and visual ap- proaches for reflective HCI design [123]. The following subsections more fully describe the ap- proach and analytical methods used in this as-yet unpublished study.

4.4.2 Participants and recruitment

We invited 341 SBTF volunteers who took part in at least one activation between 2017–2019 to complete an eligibility survey on the Qualtrics platform. The time frame delimiter was deliberate in that it reflected the period when SBTF had completed its transition from using Skype as its communication platform in 2017 to its present-day use of Slack. The volunteers were contacted via a brief email that described the study and contained a link to an online consent form and a six-question demographic survey in order to form balanced, regional participant teams, e.g., pre- ferred name and pronouns, age range, country of residence, time zone, and prior activation expe- rience. Forty-eight volunteers from North America, Europe, Africa, Asia, and Oceania represent- ing a mix of relatively new and long-term volunteers responded to the invitation and assented to the study. However, only 17 volunteers ultimately participated in the study due in large part to the growing COVID-19 pandemic and the unexpected difficulty for the regional teams to tem- porally coordinate to find a convenient time for their members to meet for the two workshops.

The range of prior activation experience spanned from 1 to over 10; with a median range of 7–9 activations for the volunteers who eventually took part in the study. So, this was a relatively ex- 89 perienced participant cohort, overall.

4.4.3 A novel adaption of participatory design for virtual teams

Previous work by Gumm [54] and Gumm et al. [55] describe a distributed PD approach to over- come physical, organizational, and temporal dimensions that prevent non-collocated teams from taking part in a participatory design process. The authors present a series of asynchronous and synchronous practices for gathering design requirements through online communication channels

(e.g., web surveys, user support emails, and annotated case studies) along with bridging activi- ties to provide context across different user groups (e.g., mediated feedback from key informants and face-to-face user/developer workshops). However, in the ensuing years since this distributed

PD method was introduced, online sociotechnical tools, like free collaborative, cloud-based infor- mation and communication technologies (ICTs) and social media platforms, have now become widely available, which temper many of the dimensional challenges that Gumm et al. needed to mitigate.

Rather than simply update the distributed PD process with more contemporary synchronous technologies, the vPD protocol differentiates itself by also addressing recent criticism that PD has strayed from its political roots [14]. The vPD workshops deliberately embed opportunities for the participants to interrogate implicit and explicit power structures—within SBTF as an organi- zation, relative to its activation partners, and/or with respect to the “algorithmic black box” of sociotechnical tools that the group appropriates for its work and of which they have little to no control over.

Building on insights about distributed temporal coordination from previous research (see Chap- ters 2/3) and my own knowledge of SBTF culture, we developed the vPD protocol for the so- ciotemporal design study to address three overarching concerns to an online site of study: 1) a synchronous, real-time format using the free, cloud-based tools the teams were already familiar with or could learn quickly with little training; 2) a setting that empowers the highly collabo- rative and values-based nature of the group; 3) flexibility to support how each team organically 90 preferred to interact with each other. To accommodate these needs, the vPD protocol uses typi- cal SBTF workflow tools: private team Slack channels as the primary communication mechanism during the workshops, Google Drive as an artifact repository, and Google Forms to collect post- workshop feedback. We also introduced Mural2, an online collaborative whiteboard platform to augment the chat discussions through free-hand sketching, virtual Post-It notes, and customized templates to support human-centered design activities, including card sorts, priority matrix, etc.

Because SBTF volunteers do not use telephonic methods for communicating with each other dur- ing activations, video conferencing tools were not considered for the protocol since they could cause technical complications for hosting the workshops as well as introduce new power dynam- ics. As it is, SBTF volunteers are quite skilled at communicating through chat-based platforms, like Slack, and this method did not appear to dissuade typical levels of conversation nor curtail their participation in the workshop activities. The vPD protocol (see Appendix A for more de- tail) was structured as two successive, 90-minute online workshops for four small teams of 3–4 participants each, conducted in the spring of 2020 (Table 4.1). Each workshop incorporated a series of team-building activities, ideation and brainstorming sessions, and opportunities for col- lective decision-making about design priorities.

Table 4.1: Study participants by SBTF organizational role and team size

Team Team A Team B Team C Team D Team E SBTF Core Team Volunteers Volunteers Core Team & Core Team & Role Volunteers Volunteers Team size 4 3 3 3 4

Workshop 1 opened with a brief introduction about the purpose and process of the study. Then, each participant read and reflected on a set of 10 Design Justice principles3 [32, 33] to empower the teams to take ownership of the vPD process and to help mitigate any power dynamics be- tween the participants and the research team. Each team discussed the principles over chat and

2https://mural.co 3https://designjustice.org/read-the-principles 91 then adopted between 3–4 as guiding values for future decisions and priorities throughout the workshops (Table 4.2). The participants completed the first workshop after brainstorming and sharing roadblocks in work practice and the appropriated sociotechnical tools that hinder their digital humanitarian work.

The second workshop also opened with a brief introduction, quickly followed by a priority-setting activity about the previously identified roadblocks. Once the priorities were determined by the team through a consensus vote, the team co-constructed potential solutions based on a collec- tively written problem statement. Throughout both workshops, the teams participated in a vari- ety of warm-up and cool down exercises to provoke creative thinking and team building, as well as provide opportunities for the participants to practice using the online whiteboard tool.

Of special note: Due to the global COVID-19 pandemic, Team A, which was composed solely of

SBTF leadership members, was delayed several weeks. In deference to the participatory design credo of ceding power to the participants, the four Core Team members ultimately decided to meet for a single 2-hour workshop that would solely focus on management issues. In the interest of time, we condensed the typical dual workshop format by eliminating the team-building activi- ties, as the members of this group all knew each other well through prior SBTF interactions.

4.4.4 Temporal priming of workshop teams

Some priming about time and temporality occurred during the workshops. Unlike Studies 1 and

2 that centered around the 2017 Hurricane Maria activation to explore temporal coordination during a particular moment in time, participants for this study were drawn from a wider array of volunteer experiences and SBTF activations. This was a deliberate recruitment strategy to evoke a broader collective vision about future work practices and appropriated technologies that would be suitable for different natural and human-caused hazards that SBTF responds to.

However, five volunteers who participated in the workshops were also previously interviewed for

Study 2, which may have resulted in team members anticipating my interest in temporal consid- erations. Further, six Core Team members participated in the workshops with whom I have had 92 prior conversations about temporal coordination problems during activations as well as sharing information with them about this research topic. Last, in contrast to the previous studies where references to time and temporality were deliberately engaged with through interview questions, discourse analysis, and theoretical work, we were interested to know if temporal concerns would emerge organically during the workshops. Therefore, only Teams B and E were specifically asked to reflect on how they made sense of social media messages that are in conflict with the relative timestamps. Due to a lack of time during the workshops, these questions were omitted from the protocol and not posed to the Teams A, C and D who thus were not primed to consider time or temporality in their brainstorming and design sessions.

4.4.5 Data corpus

The workshops generated a rich variety of text and multimedia artifacts. The corpus includes:

1,881 Slack messages (52,220 words) of dialogue from nine vPD workshop transcripts, 19 vir- tual whiteboards, 11 brief transcripts from a chat feature embedded in the whiteboard platform, and five Design Justice documents. The research team also produced its own set of artifacts: nine video screencasts of the chat streams and whiteboards, nine separate audio field notes of the live workshops, and a field note journal of post-workshop reflections. We used a talk-aloud process, common in HCI field work, to reflect in-the-moment on the workshop activities. We audio recorded the reflections and then wrote contemporaneous field notes immediately follow- ing the workshops. These notes and recordings allowed us to engage in a reflexivity process [47] to critically analyze the data and bracket observations that may have been influenced by my participant-researcher role [5, 138]. These insider insights and coded transcript excerpts were used to provide additional context about the visual artifacts produced by each of the teams. The screencasts and field notes served as referential material for understanding the context of the team dialogue and design priorities, as well as for noting insights about the workshop protocols in the moment. 93

4.4.6 Data analysis

We employed a variety of interpretative and reflective approaches to analyze the text-based and visual artifacts generated from the workshops. First, we imported each of the Slack transcripts into MaxQDA4. Two researchers read through each transcript independently and then conducted open, inductive coding and analytical memo writing [118] to interrogate the vPD artifacts within and across the teams. The researchers met to discuss differences and find collective agreement across the code themes. Next, we used affinity diagramming to identify patterns within and be- tween the teams’ whiteboarding activities. Here, we clustered the teams’ virtual Post-It notes from three online whiteboard activities: identifying roadblocks (Activity 1E), setting priorities

(Activity 2B), and collaboratively producing speculative design sketches (Activity 2E). Then, we applied a visual approach [123] to analyze the design sketches as a reflective, but unspoken dia- logue between researcher and participant to communicate meanings about the artifact [123].

4.5 Findings

4.5.1 Roadblocks to distributed temporal work

Each team used virtual Post-It notes on the whiteboard to brainstorm roadblocks, prioritize de- sign decisions based on an importance-feasibility matrix, and then sketch speculative design solu- tions. Participants were prompted during each activity to reflect on the team design principles to incorporate their collective values into the decision-making process. Being mindful of the under- lying power dynamics between the teams and myself, I was careful to not impose specific instruc- tions on how to carry out the brainstorming activity. Some teams chose to work individually and then share ideas back to the group via Slack, while others took a more collaborative approach by synchronously using the whiteboard and Slack channel to talk through ideas and disagreements to find consensus. 4https://www.maxqda.com/ 94

4.5.2 Roadblock analysis

After the fact, the research team assembled all of the virtual Post-It notes into a single white- board. We used an affinity diagramming process to interpret, compare, and rearrange the col- lective roadblocks (Figure 4.2) generated by the teams to inductively identify common themes.

We first combined like-minded ideas and then clustered the remaining 87 unique notes. Notic- ing a distinct sociotemporal thread through several of the clusters, we added red dots to indi- vidual Post-It notes to visually signify the 38 percent of roadblock brainstorms that appear to trigger a state of sociotemporal disorder in the crisis, the work, and/or the technology during an activation. For instance, participants noted a wide-range of time-related elements that further complicate their work, such as challenges in making sense of timestamp metadata, discerning the timeliness of collected data, experiencing stress about the urgency and pace of the work, ICT constraints during various phases of an activation, as well general frustrations about rework due to unclear initial briefs from activation partners, etc.

Figure 4.2: Affinity diagrams illustrating 87 unique roadblocks generated by the teams. Red dots denote a state of sociotemporal disorder.

We then used a spectrum diagram to visualize the roadblocks by category, theme, and team and 95 depict a more fine-grained view of potential areas of concern for SBTF (Figure 4.3). Four broad meta categories of roadblocks emerged from the interpretative analysis—data, technology, tasks, and work. Interleaving the clustered Post-It Notes with the team Slack channel discourse during the roadblocks activity (Figure 4.3), 18 themes come into view to provide a more detailed picture of the meta categories. While there was strong team consensus across each of the four categories, six themes garnered attention from 4 out of the 5 teams about how obstacles specifically manifest during an activation: lack of automated tools, task status confusion, getting started in the activa- tion, team communication, and work coordination.

Figure 4.3: Spectrum diagram of work practice and technology roadblocks by teams grouped by categories (center) and subthemes (perimeter).

The following excerpt5 from Team D illustrates the entangled nature of SBTF’s workflow and technology roadblocks, and why distributed temporal coordination and tool appropriation are so challenging overall for the organization:

5Volunteer names are anonymized for privacy. For continuity, volunteers are denoted both through their team membership (A, B, C, D or E) and an assigned number (e.g., A1, B3, etc.). See Table 1 for more information about the different SBTF organizational roles included in each team. 96

D2: Well, I think D1’s Issue #1,000 is a pretty big roadblock: finding a platform that works well for communication.

D1: Back in 2012 we used Skype. My notifications blew up daily with so many people talking, sharing ideas etc. But it had serious issues in loading etc so we switched to Slack. I’ve never seen the level of community in SBTF since. So So Sad

D1: Switching to Slack was good tech wise but it was the day the music died

D1: Skype had a limited number too so we had to remove people to make room for new ones and we didn’t want to do that

D3: wow @D1 curious to know more offline about the Slack dropoff. (as someone who attempts to design info solutions)

With respect to data, the volunteers shared broad concerns about misinformation on social me- dia, language translation issues, and inadequate volunteer sensemaking that called into question the trustworthiness of data sources used to curate situational awareness for the activation part- ner. A related technology roadblock was the lack of automated tools to handle repetitive data tasks more attuned to the rapid nature of the work. By defaulting to manual workarounds, the participants expressed concern about unknowingly contributing data entry errors to the acti- vation datasheet. Others noted a frustrating juxtaposition within the group: the brittleness of chaining together appropriated technologies to approximate more sophisticated data tools in con- trast to the, as yet, unmet promise of consumer-oriented machine learning tools for streaming, real-time data collection and analysis.

B2: I wish there were a platform to easily stream the newest and best imagery/layers that people of any skill level could use - not just GIS experts.

Concerns about tasks presented two intertwined obstacles for volunteers: difficulties initially get- ting started due to the high-tempo onboarding demands which drives confusion about needs, tasks, priorities, and mission status throughout the activation. Work practice challenges also 97 garnered particular attention from the teams as problems onto themselves as well as underly- ing causes of the aforementioned roadblocks. Specifically, the interplay between the overwhelming

flow of communication via Slack and the coordination fits and starts of standing up an activa- tion, often with just a few hours’ notice, were cited as particular pain points.

4.5.3 Determining design priorities

After a quick virtual whiteboard sketching exercise to prompt creative thinking, we kicked off the second workshop with a prioritization activity. The participants used an affinity diagram- ming process on the virtual whiteboard to thematically cluster the team’s roadblock brainstorms generated during the first workshop. Here, the aim was twofold: 1) to review and re-familiarize themselves with their prior work about obstacles and 2) to begin honing their initial thinking to- ward design priorities. Then, the teams placed the thematic labels on an importance-feasibility matrix (Figure 4.4) to determine which broad roadblock categories would inform a co-constructed

Figure 4.4: Prioritizing roadblocks through affinity diagramming and an Importance-Feasibility matrix.

design statement about their priorities. Last, each team member was directed to place a star icon on no more than three theme labels to further denote high level priorities for roadblocks to be 98 tackled in the forthcoming design statement and solution sketching activities.

There was surprisingly little dissent among the teams during the affinity clustering activity. This perhaps is a function of the participants’ familiarity with one another and their ease in work- ing online along with the high value placed on collaboration and achieving consensus within the broader SBTF community culture.

Across the teams, the design priorities that were rated along the high importance and high fea- sibility axis clustered between two meta-themes: 10 external technically-mediated concerns and

13 internal work practice concerns. Among the most consistent technical priorities were issues around information, including taming the volume/flow as well as developing a broader set of data sources to improve the quality and timeliness of curated information. From the standpoint of work practices, the teams were fairly unified in their desire to focus on expanding feedback to volunteers from within SBTF and by the external partners. Volunteer training and enhanced skill-task matching were also high priorities across all of the teams. One interesting temporal is- sue was raised by multiple teams about developing a flexible onboarding system that could be implemented at any juncture during an activation since volunteers may join in at various times at not necessarily at the start of the data curation project. This concern also segued into conver- sations about inter-organizational communication needs that span technology (e.g., making Slack more user-friendly) and work practices (e.g., encouraging more dialogue and camaraderie within the team).

All of the teams engaged quite deeply in the importance-feasibility exercise—discussing in detail how to define the terms and how to prioritize the roadblock themes on the matrix. The follow- ing exchange between the members of Team B occurred over 10 minutes and exemplifies the care used to understand, describe, and then apply the notion of importance to determine their priori- ties, all while moving seamlessly between Slack and the Post-It Notes on the whiteboard:

Research Team: How might we think about the “importance” of a theme? Any ideas? What does “importance” mean to you?

B1: Importance in terms of timing—needs to b done and prioritized NOW. Im- 99

portance in terms of need. Importance in terms of impact

B3: Quality of work. To direct resources to the right place on the right time. . .

B1: I think VALUE is really important..to IMPORTANCE

B3: Yes I think value is important

B1: Value might be the umbrella term that includes need, timing, impact, direc- tion

B2: if we had to prioritize those words within “importance”, I’d probably put im- pact first. In my mind, value is roughly equivalent to importance in this exercise. and timing and direction could also feed into impact.

Research Team: I LOVE this conversation :heart eyes: but I’m gonna have to nudge on us to the next activity. Sounds like importance has lots of characteris- tics

B2: hehehe. We’re chatty.

Another interesting aspect of the vPD protocol was the enthusiasm and, even, playfulness with which the teams engaged in the activities. Again, Team B illustrates both their seriousness in making very hard choices while also using humor about potential areas of contention:

Research Team: Keep in mind that we are only responsible for imagining the workflow or technology improvement not building it. We will forward the idea to the experts to make. So don’t worry too much about your own skill level to make the design work. With that in mind, how might we think about defining “feasibil- ity”?

B2: so if we aren’t building it, do we consider things like ”cost” and ”availability of current tools/software/etc” in our concept of feasibility? (edited)

Research Team: You can be broad minded about it. Skies the limit. You have no resource constraints

B2: I’m Elon Musk! 100

4.5.4 Solutions analysis

Each team produced a sketchboard with collective brainstorms on potential solutions that ad- dressed roadblocks, priorities, and incorporated the team Design Justice values (Figure 4.5).

Figure 4.5: Sketching design solutions on the virtual whiteboard.

Then, we used the same affinity diagramming process of combining, clustering, and interpreting the 122 virtual Post-It Notes generated by the teams about solutions to address the roadblocks

(Figure 4.6). We also replicated the same sociotemporal disorder meta analysis, again, depicted with red dots on the whiteboard diagram. Every team contributed at least one temporally-focused solution though they were not prompted to do so during the workshop.

Though this subset of sociotemporal solutions flowed across both technical and work themes, there was an interesting convergence among the teams around resolving temporal tensions for accomplishing tasks more effectively. These task-oriented ideas, bolstered by the ensuing team discussions, were particularly focused on developing new tools for pre-/post-activation activities, like test environments to try out new workflows or as training platforms. Likewise, there was sig- nificant agreement about the need for new techniques to better support mission-critical learning 101 that could be applied to future activations.

While most of the solutions called for the development of new tools, several ideas looked to make improvements to currently appropriated technologies. Again, focusing on concerns about so- ciotemporal disorder, the teams suggested information control flows for Slack, automating stream- ing GIS data so volunteers without experience could assist with mapping tasks, and a browser plugin to scrape and convert timestamps from social media data.

Replicating the close reading technique used to assess relationships between the roadblock brain- storms, we applied a similar interpretive analysis to the suggested solutions. We then visualized the insights through a spectrum diagram (Figure 4.7) to depict commonalities to help guide fu- ture prototype designs. Here, four distinct categories of suggested solutions emerged from the analysis—work, mission, old technology, and new technology—though there is some resonance with the roadblock categories data, technology, tasks, and work. Interleaving the clustered Post-It

Notes from the affinity diagram with the Slack channel discourse, 13 themes emerged providing additional substance for understanding the meta categories.

In a similar vein with the roadblocks analysis, there was a high level of consensus between the teams with the collective solutions. All five teams proposed at least one solution for workflow structures that would improve labor efficiencies and other ideas for developing new technical tools that would enable training/skill building during “peace time,” or the idle periods between activations. There was also an interesting concurrence between three sub-themes: communication workflows, control of information flows, and collective tasks. These concerns garnered attention from different complements of four out of five teams—including the leadership team that, at first glance in the affinity diagramming process, was understandably more preoccupied with manage- rial issues than the volunteer teams. Visualizing the collective solutions as a spectrum diagram surfaced an important insight about the leadership and volunteer experiences that was not evi- dent from the affinity mapping process.

This crossover of higher-order concerns between different levels in the SBTF hierarchy appears to indicate two possible parallel motives. One, is the common pain point of sociotemporal disorder 102

Figure 4.6: Affinity diagram of 122 unique solutions generated by the teams. Red dots denote ideas that attempt to reconcile a state of sociotemporal disorder.

Figure 4.7: Spectrum diagram of solutions suggested by teams grouped by categories (center) and subthemes. 103 that many participants raised in the brainstorming sessions, though often indirectly and not in explicit temporal language. The other likely reason is a strong commitment to SBTF and its mis- sion, as evidenced by the longstanding experience of the study participants (M = 7 activations;

Mdn = 7–9; range = 1–10+). The shared interest in resolving tensions in their work practices and appropriated technologies ultimately helps to create a more fulfilling volunteer experience.

For example, Team C approached their solution ideation process with a design statement that addresses the worth of their efforts from both internal and external points of view:

“How might we eliminate the repetitive aspects of SBTF activities and reduce the procedural mistakes to ensure SBTF volunteer work is valued and valuable?”

In their design statement, Team B took a similar stance considering insider and activation part- ner value judgments while also incorporating a desire to reconcile two forms of sociotemporal dis- order—the lack of power to manage the high-tempo nature of the data and the ephemeral tempo- ral rhythms of activation tasks:

“How might we design the SBTF volunteer experience to control information flows to provide needed information to volunteers at various stages of their work to maximize their effectiveness, and minimize their information overload?

Taming information flows through custom add-on applications to Slack channels and Google

Sheets were common refrains among the teams during their solution brainstorming sessions. In particular, developing tools to automate the overwhelming pace of data, especially from the del- uge of social media posts during a humanitarian crisis, was viewed as an opportunity to offload a high-volume, repetitive task away from volunteers with the potential benefit of reducing so- ciotemporal disorder. Likewise, there were several related suggestions to develop new workflows to capture retrospective feedback from activation partners. This is a constant frustration within the organization—something that is regularly requested by the Core Team but rarely provided to volunteers. This information, however, is highly valued by SBTF for both learning needs and training for future calls to action. The lack of feedback can also trigger sociotemporal disorder 104 when workflows and rhythms are set early in an activation but change suddenly due to the evolv- ing information needs of its partners.

Unfortunately, in this phase of the workshop, few participants attempted to sketch their specu- lative solutions despite directed prompts and planned opportunities during several team-building activities to play and practice with the virtual whiteboard drawing and image upload features.

We attribute this more to the lack of familiarity with the tools and the quick pace of the work- shops which left too little time for creative visual expression. The volunteers found it faster to simply dash out a few lines of text on a virtual Post-It Note than to illustrate their ideas.

4.6 Discussion

This study describes how globally-distributed digital humanitarian volunteers collaboratively en- visioned sociotechnical system improvements for the collection, curation, and analysis of online information generated during a mass emergency. We were especially interested in how the study participants accounted for, engaged with and/or dismissed sociotemporal disorder while reflect- ing on their high-tempo, time- and safety-critical coordination work. It comes as no surprise that designing for time in this domain remains a challenging endeavor.

Sociotemporal disorder was reflected in three distinct ways: direct interrogation of timestamps and temporal metadata that were out of sync; abstractions in team discourse; and design arti- facts generated during the workshops. These mentions were frequently interspersed with concerns about the urgent pace of the work, observations about the global distribution of the SBTF vol- unteers, concerns about overwhelming information flows, and implications for work practice and technology improvements. All of the teams engaged with multiple forms of sociotemporal disor- der. However, the design priority activities in Workshop 2 offered the most striking examples of how sociotemporal disorder can be viewed as both an important factor for achieving order as well as an agent of disorder with which the volunteers had limited ability to control.

For example, the tension between sociotemporal order and disorder was evident during a discus- sion by Team A on how to rank three high-level roadblocks: a lack of clarity from SBTF activa- 105 tion partners about their needs and a lack of post-activation feedback from the partners, both of which contribute challenges in anticipating pre-activation volunteer training needs. Here, a Core

Team member describes the influence of time on their decision about where to place the team’s roadblocks on an importance-feasibility matrix to help determine future design priorities:

A1: My feeling is that we see all three topics are crucial for the success of a de- ployment. I think they are connected to two factors: communication and time. If you know what to do, you save time. And time is crucial in these activations, as it’s very limited.

The realization that sociotemporal disorder plays an important, but often unacknowledged, role in their work was also mentioned during the Team D discussion on prioritizing roadblocks and the group’s lack of resources that forces volunteer labor appropriating free, cloud-based collabora- tion technologies over hired staff using customizable enterprise-level tools.

D1: I placed Time Constraints low on feasibility because I don’t think we can control this as a volunteer group. It’s a struggle with every Core Team since 2010. It’s not really solvable and we can’t move to a paid organization. It’s just a little problem that we have learned to accept.

4.6.1 Themes across artifacts

We identified five themes across the text-based and visual artifacts that were relevant to both

SBTF work practices and the technologies they appropriate to conduct their digital humanitarian work. The first three themes—roadblocks, priorities, and solutions—represent the interrelated steps in the design process. For example, we found that even in this new context of workshop dialogue, SBTF volunteers used the same language to communicate about time and the same strategies for navigating sociotemporal order and disorder as described in Study 1. Likewise, the teams used the same sensemaking strategies illustrated in Study 2 to wrestle with instances of sociotemporal disorder. For example, Team E chats about the impact of temporal distribution on work practices and technologies in this brief Slack excerpt of a conversation following a team- building exercise about the global nature of the SBTF community: 106

E3: I actually think working across time zones is a benefit.

E2: I agree completely.

E1: At best we cover 24 hrs of activity. . .

E3: At least in the activations that I’ve worked in, it has always helped to keep the team active around the clock. It does mean that your coordinators need to be globally dispersed, but again I haven’t experienced an issue.

E2: Big issue is when clocks go forward or back in different weeks.

E4: As a volunteer/contributor, a different timezone is not an issue in my opin- ion. It’s a challenge for the Core Team to be available and Maintain continuity. But sometimes for volunteers, even say after a 1-2 hour break, there’s so much done by others, it’s difficult to follow up

E3: Re: the post, check geo-location, and sometimes you can look at the timing of comments and responses to make sense of the validity of the post itself.

The difficulty for the participants to articulate sociotemporality was borne out in this study as it was in previous work. On the surface, most explicit temporal references invoked clocks, times- tamps, metadata, time zones, and other instrumented forms of time. Mentions of temporal ex- periences, like achieving sociotemporal order or attempting to overcome a sense of sociotemporal disorder to enable productive work, were often far more subtle and required multiple analytical passes across the Slack chat transcripts and visual artifacts. Yet, despite the lack of discursive precision there was no lack of ideas about roadblocks or solutions that had a temporal compo- nent. Drawing from pluritemporality [98] provided a way to infer more descriptive language in the team conversations and sketches about the experience of distributed temporal coordination work. But the ubiquity of this wide-ranging social construct that incorporates multiple domains of time also runs the risk of further complicating an already quite entangled form of globally dis- persed online work.

To guard against theoretical overreach here, we relied on our pluritemporal analytical framework

[93] as a way to attend to aspects of time and sociotemporal experiences that were most salient 107 to this form of work. In particular, we were able to more quickly and efficiently discern tempo- ral meanings from the Slack transcripts and whiteboard activities by relying on examples of the pluritemporal references to time, such as a standard format (e.g., time stamps), interval (e.g., timespans), chronology (e.g., linear time), tempo (e.g., pace/flow), presence (e.g., metaphorical spatial references), and ephemerality (e.g., fleeting nature) during the affinity diagramming pro- cess rather than re-interrogating temporal language usage from scratch.

The pluritemporal analytical framework was especially helpful for discerning in what ways those temporal shifts might occur. This was particularly evident when several teams wrestled with in- terrelated issues that drift between sociotemporal order to disorder and then back to some sem- blance of order. Here, we also found articulations about time that could be incorporated into the design of potential solutions as technical affordances, smart prompts, or predictive analytics.

However, the discursive nature of sociotemporality is so subtle that it demands close observation to pick up on it—well beyond the limits of everyday people engaged in a codesign process for a sociotechnical system. Building out practical applications for the pluritemporal analytical frame- work in HCI design work will be incorporated into future work.

4.7 Design Implications and Future Work

The dichotomy between sociotemporal order and a sense of control is substantially rooted in the multifaceted ways that the group experiences disorder—in the crisis, in the work coordination, and in the sociotechnical tools, as described in Chapter 2. The seemingly straightforward design- focused research question in this study—in what ways might collaborative technologies better sup- port the temporal coordination work of globally distributed teams?—reveals a much more profound problem upon analyzing the data. Are we designing to repair the disorder to wrest control of it or to embrace disorder as a resource for design? In short, how do we approach these sociotempo- ral realities: as improvisational opportunities [65], creative reconfigurations of plans and situated actions [134], existential threats to safety-critical work [146], or something in-between?

Prior work in CSCW, HCI, and adjacent topics have explored related “X as a resource” concepts 108 for design, such as ambiguity [48], friction [137], context [70], seams [22], Inman2019, rather than as a problem to overcome. Though temporality is often mentioned in this literature, none of it expressly considers designing for time. In future work, we will build on the previous chapters to consider sociotemporal disorder as a resource in the design of sociotechnical systems.

4.8 Conclusion

This research describes insights about designing sociotechnical systems that can better support the temporal coordination of SBTF, a globally-dispersed digital humanitarian organization, that is engaged in high tempo, time- and safety-critical work. We used a novel virtual participatory design (vPD) approach augmented by values-based design justice principles to deeply interrogate roadblocks, priorities, and potential solutions with five teams using familiar tools and techniques equivalent to their typical online-only, chat-driven work practices and organizational culture. The vPD protocol was deliberately crafted to provide a useful co-design template for online teams to provoke opportunities to confront the power dynamics of new configurations of labor online and the free, commercial technologies used in service of that work. This research contributes several empirical insights about the ways in which sociotemporal disorder manifests: as a lack of collec- tive agreement about time that cause discordant temporal experiences; the deep entanglements between ad hoc work practices and appropriated sociotechnical systems in this context that drive sociotemporal disorder; and the capacity for creative approaches to surface high-level abstrac- tions, like sociotemporality, that are often invisible in discourse. 109

Table 4.2: Activity protocol for each vPD workshop.

Activity / Time Description Workshop 1: Team brainstorming and design value decisions Activity 1A Welcome and introductions of the research team and partici- (5 minutes) pants. Reiteration of study purpose, objectives, and consent con- ditions. Activity 1B Warmup exercise designed to build inter-team camaraderie. (10 minutes) Activity 1C Team develops a set of shared design values to guide ideation (30 minutes) and prototyping activities to empower participants to claim ownership of the PD process and mitigate power dynamics be- tween the teams and researcher. A Design Justice zine [31] will be pinned to the Slack channel for inspiration/adaptation. Activity 1D Cool down exercise to build inter-team camaraderie and practice (10 minutes) collaborative sketching using the online whiteboard. Activity 1E Temporal brainstorming activity on a whiteboard. Sub-teams (30 minutes) will use archival probes to identify how time is represented in SBTF work practices, technologies, and artifacts. Activity 1F Team debrief in Slack to conclude workshop 1 to share thoughts, (5 minutes) recalibrate goals, etc. A link to an anonymous survey will also be distributed to the participants to solicit procedural feedback. Workshop 2: Team co-design and prototyping priorities Activity 2A Warmup exercise to prompt creative thinking. (5 minutes) Activity 2B Sub-teams review whiteboard brainstorms, discuss/debate ideas (20 minutes) on Slack, continue brainstorming (if desired), and vote on priori- ties. Activity 2C Team reconvenes to present, negotiate, and prioritize sub-team (15 minutes) whiteboard sketches. Activity 2D Cool down exercise to encourage creative, divergent thinking (5 minutes) among teams. Activity 2E Full team activity to discuss, sketch, improvise, re-prioritize, en- (30 minutes) vision, and prototype design mockups of priority technology im- provements Activity 2F Discussion of research integration of the team’s final design (5 minutes) mockups and next steps. Activity 2G Final debrief in Slack to conclude study, to share thoughts, and (5 minutes) ask lingering questions. A link to an anonymous survey will also be distributed to the participants to solicit procedural feedback. Chapter 5

Designing for Sociotemporal Disorder

CHAPTER 5: Discussion and Design Implications

This multidisciplinary, multi-methods research explored how sociotemporality is implicated in globally distributed digital humanitarian work discursively, in sensemaking, and as a design re- source. Building on the work of social theorists, Adam (timescapes), Nowotny (pluritemporality), and Zerubavel (temporal language and semiotics), this dissertation seeks to contribute an ap- plied lens to the study of temporality: how time is socially structured and experienced in work

(Chapter 2); strategies for collectively discerning high-tempo information within a team (Chapter

3); and the ways in which time, and sociotemporal disorder in particular, could be better inte- grated into technologies used for collaborative work (Chapter 4). Together, this empirical work contributes a roadmap to address a longstanding gap in the CSCW and HCI literature by inter- leaving the social experience of temporality with instrumented time in sociotechnical systems.

The focus of Chapter 2 takes both empirical and theoretical approaches for building deeper un- derstandings about how a distributed team collectively uses language to allow their temporal coordination work to productively unfold. Here, we used a variety of text and visual artifacts generated by SBTF during the 2017 Hurricane Maria in Puerto Rico activation to explore the concept of sociotemporality in an applied setting. In the process of developing a pluritemporal analytical framework to interrogate the discursive aspects of this high-tempo temporal coordina- tion, I came to realize how challenging it is for distributed teams to achieve a sense of order, or collective agreement about their temporal experiences. From this discovery and subsequent ex- 111 amination of discordant examples of distributed online coordination gone wrong in the artifacts, my coauthors and I began to explore a new theoretical concept: sociotemporal disorder, or a breakdown in putting collective agreement about time into action.

This description appears to be the first reference to the construct of temporal misalignment be- tween the social experience of time and its instrumented representation that hinders coordina- tion. We concluded this chapter by describing a set of design implications that take into account the variety of ways that sociotemporal disorder was revealed in the empirical analysis and offered speculative ideas on how to potentially resolve the problems. At this point in the overarching re- search agenda, the broader research question was to understand how sociotemporal disorder in digital humanitarian work might be reconciled to enable more productive distributed work.

Hence, the interview study in Chapter 3 focused on how the team engaged in collective sense- making, again using the Hurricane Maria activation as the research site. In true digital humani- tarian optimistic fashion, I embarked on this research believing that the SBTF volunteers could easily relate to the sociotemporal tensions that I had revealed in the initial artifact study since we had collectively experienced these problems during the activation. I could not have been more wrong.

After two pilot interviews with highly experienced volunteers, it became glaringly obvious that it was very challenging to articulate their temporal experiences. Asking direct questions about sociotemporal disorder did not result in the rich dialogue I had hoped for and expected. Instead, the volunteers struggled to talk about time and temporality with any specificity. For instance, the conversations often stalled beyond simple sharing frustrations about certain tools, such as the lack of fidelity in Google Sheets to enforce custom date- and timestamp format settings when dozens of volunteers were entering information. Consequently, I scrapped the initial interview protocol and regrouped by designing a new set of questions that were paired with artifact probes, e.g., transcript excerpts, social media posts, and maps that exemplified specific temporal coordi- nation problems during the activation. These memory indices helped to recenter the discussion and elicited more insightful responses. A co-author also suggested incorporating a question that 112 would allow the interview participants an opportunity to envision new work practices or tech- nologies to help alleviate the strain caused by sociotemporal disordered encounters during the activation. Again, this more creative and open-ended interview approach, in contrast to the ear- lier journalistic style, evoked interesting insights about how the SBTF volunteers were adopting different sensemaking strategies in their work.

In analyzing the interview data from this study, I was struck by a quote from one of the partici- pants: “People talk in stories, but response teams talk in spreadsheets, charts, and maps.” Here, she describes the disconnect between the raw data of social media posted by everyday people about the crisis in contrast to the structured data output expected by the response profession- als.

The final dissertation study was designed to put into practice the insights derived from the prior research. As I found in the theoretical work from the SBTF Hurricane Maria content analysis study reprinted in Chapter 2 and the interview study reprinted in Chapter 3, the ways in which people organize shared understandings of time is both understudied and offers intriguing chal- lenges for the design of sociotechnical systems used in distributed temporal coordination work.

Further exploration of these parallel and often quite entangled sociotemporal experiences and their design implications, motivate the research described in Chapter 4.

This inability to reconcile sociotemporal disorder continues to spur work practice and technolog- ical frictions for SBTF that have yet to be resolved despite numerous makeshift approaches and manual workarounds. Again, as I discovered in the interview study on the discursive challenges of talking about time, abstractions, like sociotemporal dis/order, are also not concepts that can be effectively addressed head-on with the participants. While the decision to host virtual participa- tory design workshops was a big gambit (and could have easily fallen flat), the chance to deeply engage with the SBTF volunteers in a new way to extend the insights from the prior ethnographic work, content analysis, and interviews felt like a worthwhile risk.

An unexpected outcome of this research was a dawning realization that designing new work prac- tices or sociotechnical tools to reconcile sociotemporal disorder may not be the optimal approach. 113

Rather than play a whack-a-mole trying to anticipate every potential instance of sociotemporal disorder that the team might encounter, I became intrigued by the idea that taking advantage of the adaptable culture of SBTF offers a more fruitful approach. This led me to explore how to interpret the ideas generated by the workshop teams as spontaneous opportunities to learn from and adapt to the sociotemporal disorder of the moment. Klemp et al. [65] describe how a jazz ensemble temporally recovers from a bad note as an example of improvisational learning that provides an interesting new perspective on how to think about incorporating something unex- pected as creative grist rather than as a problem to solve:

“A wrong note is errant only to what has already happened, and it can be made less errant by rearranging what happens next. A mis-take is in this way a spon- taneous move in a system of moves in search of connections that carry forward” (p. 10).

5.0.1 New interpretations of speculative designs

With this new perspective of sociotemporal disorder as a resource for design, I am experimenting with how this idea for future work might influence the speculative ideas that the SBTF volun- teers offered in the interviews and participatory design workshops. One approach is to improve

Tweetdeck, a social media management platform commonly used by SBTF volunteers during an activation to monitor a variety of Twitter keywords, hashtags, and user accounts. In this illus- tration (Figure 5.1), the user is granted new affordances to control how timestamp and timezone metadata are represented in a tweet about Hurricane Maria through a simple add-on tool.

Rendering sociotemporal disorder as visible enables a map user to easily detect trends in the age, number, and location of data points (Figure 5.2). This interpretation of a data dashboard idea raised in both the interview study and participatory design workshops also addresses a leadership team desire to automate data monitoring tasks. Here, the opportunity to center sociotemporal disorder as a normal and expected occurrence during a crisis could lend more confidence in the data products curated by SBTF for emergency responders. 114

Figure 5.1: Tweetdeck add-on prototype that affords users control over temporal metadata.

Figure 5.2: AI-enabled heat map that reflects trends in the number of reports and the age of data collected during an activation. Bibliography

[1] Barbara Adam. Social versus natural time, a traditional distinction re-examined. In M Young and T Schuller, editors, The Rhythms of Society. Routledge, London, UK, 1988.

[2] Barbara Adam. Time and Social Theory. John Wiley & Sons, 1990.

[3] Barbara Adam. Beyond boundaries: Reconceptualizing time in the face of global chal- lenges. Social Science Information, 33(4):597–620, 1994.

[4] Barbara Adam. The Quest for Time Control. In Time, chapter 6, pages 123–148. Polity Press, Cambridge, UK, 2004.

[5] Kathryn J. Ahern. Pearls, Pith, and Provocation: Ten Tips for Reflexive Bracketing. Qualitative Health Research, 9(3):407–411, 1999.

[6] Jennings Anderson, Robert Soden, Brian Keegan, Leysia Palen, and Kenneth M. Anderson. The crowd is the territory: Assessing quality in peer-produced spatial data during disas- ters. International Journal of Human-Computer Interaction, 34(4):295–310, 2018.

[7] Jakob E. Bardram. Temporal Coordination: On Time and Coordination of Collaborative Activities at a Surgical Department. Computer Supported Cooperative Work, 9:157–187, 2000.

[8] Jakob E. Bardram and Thomas R. Hansen. Context-Based Workplace Awareness. Computer Supported Cooperative Work, 19(2):105–138, 2010.

[9] Jeanette Blomberg and Helena Karasti. Reflections on 25 years of ethnography in CSCW. Computer Supported Cooperative Work: CSCW: An International Journal, 22(4–6):373– 423, 2013.

[10] Allen C. Bluedorn. The human organization of time: Temporal realities and experience. Stanford University Press, Stanford, CA, 2002.

[11] Sara A. Bly, Steve R. Harrison, and Susan Irwin. Media spaces: Bringing people together in a video, audio, and computing environment. Communications of the ACM, 36(1):28–46, 1993.

[12] Susanne Bødker, Pelle Ehn, Dan Sj¨ogren,and Yngve Sundblad. Cooperative design per- spectives on 20 years with ‘The Scandinavian IT Design Model’. In Proceedings of the first Nordic Conference on Human-Computer Interaction, 2000. 116

[13] Susanne Bødker, Kaj Grøenbæk, and Morten Kyng. Cooperative Design: Techniques and Experiences from the Scandinavian Scene. In D. Schuler and A. Namioka, editors, Participatory Design: Principles and Practices, pages 157–175. Lawrence Erlbaum Asso- ciates, 1993.

[14] Susanne Bødker and Morten Kyng. Participatory Design that Matters—Facing the Big Issues. ACM Transactions on Computer-Human Interaction, 25(1):1–31, 2018.

[15] David M Boje. The Storytelling Organization : A Study of Story Performance in an Office- Supply Firm. Administrative Science Quarterly, 36(1):106–126, 1991.

[16] Lera Boroditsky. Metaphoric structuring: Understanding time through spatial metaphors. Cognition, 75(1):1–28, 2000.

[17] Tone Bratteteig and Ina Wagner. Unpacking the notion of participation in participatory design. Computer Supported Cooperative Work, 25(6):425–475, 2016.

[18] Ryan Burns. Rethinking big data in digital humanitarianism: practices, epistemologies, and social relations. GeoJournal, 80:477–490, 2015.

[19] Erran Carmel, Randall D. Whitaker, and Joey F. George. PD and Joint Application De- sign: A Transatlantic Comparison. Communications of the ACM, 36(4):40–48, 1993.

[20] Centre for Research on the Epidemiology of Disasters. Emergency Events Database (EM- DAT), 2019.

[21] Centre for Research on the Epidemiology of Disasters (CRED) and The United Nations Office for Disaster Risk Reduction (UNISDR). The Human Cost of Weather Related Disasters 1995-2015. Brussels, Belgium, 2015.

[22] Mark Chalmers, Ian MacColl, and Marek Bell. Proceedings of Eurowearable 2003.

[23] J. I. Charlton. Nothing About Us Without Us: Disability, oppression and empowerment. University of California Press, 1998.

[24] Kathy Charmaz. Constructing grounded theory. Sage, 2014.

[25] Lars Rune Christensen, Hasib Ahsan, and Eshrat Akand. {K}rishi {K}ontho: {A}n Agri- cultural Information Service in Bangladesh. In Proceedings of the 10th Nordic Conference on Human-Computer Interaction, NordiCHI ’18, pages 203–214. ACM, 2018.

[26] Herbert H. Clark and Susan E. Brennan. Grounding in Communication. In L.B. Resnick, J. Levine, and S. Teasley, editors, Perspectives on Socially Shared Cognition, pages 127– 149. APA Press, 1991.

[27] Andrew Clement. Computing at work: empowering action by low-level users. In Computerization and Controversy-Value Conflicts and Social Choices, pages 383–406. Mor- gan Kaufmann, 1996.

[28] Camille Cobb, Ted Mccarthy, Annuska Perkins, Ankitha Bharadwaj, Jared Comis, Brian Do, and Kate Starbird. Designing for the Deluge: Understanding & Supporting the Dis- tributed, Collaborative Work of Crisis Volunteers. In CSCW 2014, pages 888–899, 2014. 117

[29] Tina Comes, Olga Vybornova, and Bartel Van de Walle. Structured Data for Humanitarian Technologies: Perfect Fit or Overkill? In Papers from the AAAI Spring Symposium, pages 7–11. AAAI Press, 2015.

[30] Commonwealth of Puerto Rico. Transformation and Innovation in the Wake of Disaster: An Economic and Disaster Recovery Plan for Puerto Rico. Technical report, 2018.

[31] Sasha Costanza-Chock. Issue #3: Design Justice for Action, 2017.

[32] Sasha Costanza-Chock. Design Justice: towards an intersectional feminist framework for design theory and practice. DRS2018: Catalyst, 2, 2018.

[33] Sasha Costanza-Chock. Design Justice: Community-led practices to build the worlds we need. MIT Press, 2020.

[34] Kate Crawford and Megan Finn. The limits of crisis data: analytical and ethical challenges of using social and mobile data to understand disasters. GeoJournal, 80(4):491–502, 2014.

[35] Kevin Crowston and Ericka Eve Kammerer. Coordination and collective mind in software requirements development. IBM Systems Journal, 37(2):227–245, 1998.

[36] Ann L. Cunliffe and Chris Coupland. From hero to villain to hero: Making experience sen- sible through embodied narrative sensemaking. Human Relations, 65(1):63–88, 2012.

[37] Ann L. Cunliffe, John T. Luhman, and David M. Boje. Narrative Temporality: Implica- tions for Organizational Research. Organization Studies, 25(2):261–286, 2004.

[38] Patrick Dawson and Christopher Sykes. Concepts of Time and Temporality in the Sto- rytelling and Sensemaking Literatures: A Review and Critique. International Journal of Management Reviews, 00:1–18, 2018.

[39] Leon Derczynski, Kenny Meesters, Kalina Bontcheva, and Diana Maynard. Helping Crisis Responders Find the Informative Needle in the Tweet Haystack. In Proceedings of the 15th ISCRAM Conference, Rochester, NY, 2018.

[40] Paul Dourish and Victoria Bellotti. Awareness and coordination in shared workspaces. In Proceedings of the 1992 ACM conference on Computer-supported cooperative work, pages 107–114, 1992.

[41] Paul Drew. When Documents ’Speak’: Documents , Language and Interaction. In Paul Drew, Geoffrey Raymond, and Darin Weinberg, editors, Talk and Interaction in Social Research Methods, pages 63–80. SAGE Publications Ltd, 2011.

[42] Edeltraud Egger and Ina Wagner. Negotiating temporal orders. Computer Supported Cooperative Work (CSCW), 1(4):255–275, 1992.

[43] Yrj¨oEngestr¨om. Learning, Working and Imagining. Twelve Studies in Activity Theory. Orienta-Konsultit Oy, 1990.

[44] Ingrid Erickson and Melissa Mazmanian. Bending Time to a New End: Investigating the Idea of Temporal Entrepreneurship. In Judy Wajcman and Nigel Dodd, editors, The Sociology of Speed: Digital, Organizational and Social Temporalities, chapter 11, pages 1– 27. Oxford University Press, 2016. 118

[45] T Erickson and W A Kellogg. Social Translucence: An Approach to Designing Systems that Support Social Processes. ACM Transactions on Computer-Human Interaction, 7(1):59–83, 2000.

[46] Samer Faraj and Yan Xiao. Coordination in Fast-Response Organizations. Management Science, 52(8):1155–1169, 2006.

[47] Jan Fook, Val Collington, Fiona Ross, Gillian Ruch, and Linden West. Researching Critical Reflection: Multidisciplinary Perspectives. Taylor & Francis, 2015.

[48] William Gaver, Jacob Beaver, and Steve Benford. Ambiguity as a resource for design. Conference on Human Factors in Computing Systems - Proceedings, pages 233–240, 2003.

[49] David Gewirtz. Technology that Changed Us: The 2000s, from iPhone to Twitter. ZDNet, 2018.

[50] Anthony Giddens. The Consequences of Modernity. Stanford University Press, 1990.

[51] Paul Glennie and Nigel Thrift. Revolutions in the Times: Clocks and the temporal struc- tures of everyday life. In D.N. Livingston and C.W.J. Withers, editors, Geography and Revolution. University of Chicago Press, Chicago, 2005.

[52] Erica Gralla, Jarrod Goentzel, and B V de Walle. Understanding the information needs of field-based decision-makers in humanitarian response to sudden onset disasters. In Leysia Palen, Monika B¨uscher, Tina Comes, and Amanda Hughes, editors, Proceedings of the 12th International Conference on Information Systems for Crisis Response and Management (ISCRAM), pages 1–7, Kristiansand, Norway, 2015. ISCRAM.

[53] Elizabeth Grosz. The Nick of Time: Politics, evolution, and the untimely. Duke University Press, 2004.

[54] Dorina C Gumm. Distributed participatory design: An inherent paradoxon? Proceedings of IRIS29, pages 1–17, 2006.

[55] Dorina C Gumm, Monique Janneck, and Matthias Finck. Distributed Participatory Design – A Case Study. NordCHI 2006, pages 1–5, 2006.

[56] Edward T. Hall. The Dance of Life: The Other Dimension of Time. Anchor Press, 1983.

[57] Gillian R Hayes. Design, Action, and Practice: Three Branches of the Same Tree. In Volker Wulf, Volkmar Pipek, David Randall, Markus Rohde, Kjeld Schmidt, and Gunnar Stevens, editors, Socio-Informatics, pages 303–318. Oxford University Press, 2018.

[58] Daniel E. Hellmann, Carleen F. Maitland, and Andrea H. Tapia. Collaborative Analyt- ics and Brokering in Digital Humanitarian Response. In Proceedings of the 19th ACM Conference on Computer-Supported Cooperative Work & Social Computing - CSCW ’16, pages 1282–1292, 2016.

[59] Alan R. Hevner, Salvatore T. March, Jinsoo Park, and Sugha Ram. Design Science in In- formation Systems Research. MIS Quarterly, 28(1):75–105, 2004. 119

[60] Stephanie Houde and Charles Hill. What do Prototypes Prototype? Handbook of Human-Computer Interaction, pages 367–381, 1997.

[61] Amanda Lee Hughes and Andrea H. Tapia. Social Media in Crisis When Professional Responders Meet Digital Volunteers. Homeland Security & Emergency Management, 12(3):679–706, 2015.

[62] Margareta Hult and Sven-Ake˚ Lennung. Towards a Definition of Action Research: A Note and Bibliography. Journal of Management Studies, 17(2):241–250, 1980.

[63] Sarah Inman and David Ribes. “Beautiful seams”: Strategic revelations and concealments. In Conference on Human Factors in Computing Systems - Proceedings, pages 1–13, 2019.

[64] International Telecommunications Union. Measuring Digital Development: Facts and Fig- ures, 2019.

[65] Nathaniel Klemp, Ray McDermott, Jason Duque, Matthew Thibeault, Kimberly Powell, and Daniel J. Levitin. Plans, takes, and mis-takes. Critical Practice Studies, 10(1):4–21, 2008.

[66] Sarah Kuhn and Michael J. Muller. Participatory Design. Communications of the ACM, 36(4):24–28, 1993.

[67] Diana S. Kusunoki and Aleksandra Sarcevic. Designing for Temporal Awareness: The Role of Temporality in Time-Critical Medical Teamwork. In CSCW ’15, pages 1465–1476, Van- couver, BC, Canada, 2015.

[68] George Lakoff and Mark Johnson. Metaphors We Live By. The University of Chicago Press, Chicago, IL, USA, 1998.

[69] Jean Lave and Etienne Wenger. Situated learning: Legitimate peripheral participation. Cambridge University Press, 1991.

[70] Christopher A. Le Dantec. Situating design as social creation and cultural cognition. CoDesign, 6(4):207–224, 2010.

[71] J David Lewis and Andrew J Weigert. The Structures and Meanings of Social Time. Social Forces, 60(2):432–462, 1981.

[72] Yvonna S Lincoln and Egon G Guba. Establishing trustworthiness. In Naturalistic Inquiry, chapter 11, pages 289–327. SAGE, Newbury Park, CA, 1985.

[73] SiˆanE. Lindley. Making Time. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing - CSCW ’15, pages 1442–1452, 2015.

[74] Rich Ling. The Mobile Connection: The cell phone’s impact on society. Morgan Kaufmann, 2004.

[75] Rich Ling and Birgitte Yttri. Hyper-coordination via mobile phones in Norway. In J Katz and M Aakhus, editors, Perpetual contact: Mobile communication, private talk, public performance, pages 139–169. Cambridge University Press, Cambridge, UK, 2002. 120

[76] Thomas Ludwig, Christian Reuter, and Volkmar Pipek. Social Haystack: Dynamic Qual- ity Assessment of Citizen-Generated Content during Emergencies. ACM Transactions on Computer-Human Interaction (TOCHI), 22(4):1–27, 2015.

[77] Mirca Madianou. Digital Inequality and Second-Order Disasters: Social Media in the Ty- phoon Haiyan Recovery. Social Media + Society, 1(2), 2015.

[78] Lena Mamykina, Elizabeth Mynatt, and Michael A Terry. Time Aura: Interfaces for Pac- ing. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI ’01, pages 144–151. ACM, 2001.

[79] Catherine C. Marshall and Frank M. Shipman. Spatial hypertext and the practice of infor- mation triage. Proceedings of the ACM Conference on Hypertext, pages 124–133, 1997.

[80] Thomas L Martin. Time and time again: Parallels in the development of the watch and the wearable computer. In Proceedings of the 2002 International Symposium on Wearable Computers, ISWC 2002, pages 5–11, 2002.

[81] Charles Martin-Shields. The Technologists Dilemma: Ethical Challenges of Using Crowd- sourcing Technology in Conflict and Disaster-Affected Regions. Georgetown Journal of International Affairs, 14:157–163, 2013.

[82] Melissa Mazmanian, Ingrid Erickson, and Ellie Harmon. Circumscribed Time and Porous Time. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing - CSCW ’15, pages 1453–1464, 2015.

[83] Joseph E. McGrath and Janice R. Kelly. Time and Human Interaction: Toward a Social Psychology of Time. Guilford Press, 1986.

[84] Patrick Meier. Digital Humanitarians: How big data is changing the face of humanitarian response. CRC Press, 2015.

[85] Naja L Holten Møller and Pernille Bjørn. In Due Time: Decision-Making in Architectural Design of Hospitals. In COOP 2016: Proceedings of the 12th International Conference on the Design of Cooperative Systems, pages 191–206. Springer International, 2016.

[86] Michael Muller and Allison Druin. Participatory Design : The Third Space in HCI. In Human Computer Interaction Handbook, pages 1125–1153. CRC Press, 2012.

[87] Michael J Muller, Werner Geyer, Beth Brownholtz, Eric Wilcox, and David R Millen. One- hundred days in an activity-centric collaboration environment based on shared objects. In Proceedings of the {SIGCHI} conference on {Human} factors in computing systems, pages 375–382. ACM, 2004.

[88] Nancy D Munn. The Cultural Anthropology of Time : A Critical Essay. Annual Review of Anthropology, 21:93–123, 1992.

[89] Bonnie A Nardi. Activity Theory and Human-Computer Interaction. In B.A. Nardi, editor, Context and Consciousness: Activity Theory and Human-Computer Interaction. The MIT Press., 1996. 121

[90] Bonnie A Nardi, James R Miller, and David J Wright. Collaborative, Programmable Intel- ligent Agents. Communications of the ACM, 41(3):96–104, Mar 1998.

[91] David M Neal. Social Time and Disaster. International Journal of Mass Emergencies & Disasters, 31(2):247–270, 2013.

[92] Wendy Norris. Digital Humanitarians: Citizen journalists on the virtual front line of natu- ral and human-caused disasters. Journalism Practice, 11(2-3):213–228, 2017.

[93] Wendy Norris, Amy Voida, Leysia Palen, and Stephen Voida. ‘Is the time right now?’: Reconciling sociotemporal disorder in distributed team work. Proceedings of the ACM on Human-Computer Interaction, 3(CSCW), 2019.

[94] Wendy Norris and Stephen Voida. Models and Metaphors of Temporality: Support- ing individual- and group-based time-management and coordination work. In Position paper for the Symposium on HCI Across Borders, held in conjunction with the SIGCHI Conference on Human Factors in Computing Systems (CHI 2017), Denver, CO, 2017.

[95] Wendy Norris and Stephen Voida. Temporality in Crisis Informatics: Representations and integrations of time in humanitarian crowd work. In Grand Challenges for HCI Crisis Informatics Researchers Workshop, 2017.

[96] Wendy Norris and Stephen Voida. Temporality in Crisis Informatics: Representations of time in digital humanitarian systems. In Position paper for the Theory Transfers? Social Theory and CSCW Research Workshop, held in conjunction with the ACM Conference on Computer-Supported Cooperative Work (CSCW 2017), Portland, OR, 2017.

[97] Wendy Norris, Stephen Voida, Leysia A Palen, and Kenneth Mark Anderson. Unwinding Pluritemporal Time in Digital Humanitarian Crowdwork. In CHI 2019 Workshop on the Future of Work, 2019.

[98] Helga Nowotny. Time and Social Theory: Towards a Social Theory of Time”. Time & Society, 1(3):421–454, 1992.

[99] William Odom, Richard Banks, Abigail Durrant, David Kirk, and James Pierce. Slow tech- nology. In Proceedings of the Designing Interactive Systems Conference on - DIS ’12, pages 816–817. ACM Press, 2012.

[100] Wanda J. Orlikowski and JoAnne Yates. It’s About Time: Temporal Structuring in Orga- nizations. Organization Science, 13(6):684–700, 2002.

[101] Leysia Palen and Kenneth Mark Anderson. Crisis informatics: New data for extraordinary times. Science, 353(6296):224–225, 2016.

[102] Leysia Palen, Sarah Vieweg, and Kenneth Mark Anderson. Supporting Everyday Ana- lysts in Safety- and Time-Critical Situations. The Information Society: An International Journal, 27(1):52–62, 2010.

[103] Michael Quinn Patton. Qualitative Research & Evaluation Methods: Integrating theory and practice. SAGE, 4th edition, 2015. 122

[104] James W. Pennebaker and Kent D. Harber. A Social Stage Model of Collective Coping: The Loma Prieta Earthquake and The Persian Gulf War. Journal of Social Issues, 1993.

[105] Nancy Pennington and Reid Hastie. A theory of explanation-based decision making. In G. Klein, J. Orasanu, R. Calderwood, and C.E. Zsambok, editors, Decision Making in Action: Models and methods, pages 188–204. Ablex, 1993.

[106] Volkmar Pipek, Sophia B. Liu, and Andruid Kerne. Crisis Informatics and Collaboration: A Brief Introduction. Computer Supported Cooperative Work, 23:339–345, 2014.

[107] Peter Pirolli and Daniel M. Russell. Introduction to This Special Issue on Sensemaking. Human-Computer Interaction, 26(1-2):1–8, 2011.

[108] Larissa Pschetz, Michelle Bastian, and Chris Speed. Temporal design: Looking at time as social coordination. In Design + Research + Society, Brighton, UK, 2016.

[109] E.L. Quarantelli. What is a disaster?: A dozen perspectives on the question. Routledge, 2005.

[110] Emily Rader and Rebecca Gray. Understanding user beliefs about algorithmic curation in the Facebook news feed. In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, pages 173–182, 2015.

[111] R´ois´ınRead, Bertrand Taithe, and Roger MacGinty. Data Hubris - Humanitarian Informa- tion Systems and the Mirage of Technology. Third World Quarterly, pages 1–18, 2016.

[112] Madhu C. Reddy, Paul Dourish, and Wanda Pratt. Temporality in medical work: Time also matters. Computer Supported Cooperative Work, 15(1):29–53, 2006.

[113] Juliane Reinecke and Shahzad Ansari. Time, Temporality, and Process Studies. In Ann Langley and Hari Tsoukas, editors, The SAGE Handbook of Process Organization Studies, chapter 25, pages 303–320. Sage, 2016.

[114] Markus Rohde and Volker Wulf. Integrated Organization and Technology Development: A Critical Evaluation. In G. Wulf, V., Pipek, V., Randall, D., Rohde, M., Schmidt, K. and Stevens, editor, Socio-Informatics. Oxford University Press, 2018.

[115] Ragnar Rommetveit. On the Architecture of Intersubjectivity. In L.H. Strickland, F.E. Aboud, and K.J. Gergen, editors, Social Psychology in Transition, pages 201–214. Springer, 1976.

[116] Joanna Saad-Sulonen, Eva Eriksson, Kim Halskov, Helena Karasti, and John Vines. Un- folding participation over time: temporal lenses in participatory design. CoDesign, 14(1):4– 16, 2018.

[117] Sundeep Sahay. Implementation of information technology: a space-time perspective. Organisational Studies, 18(2):229–260, 1997.

[118] Johnny Salda˜na. The Coding Manual for Qualitative Researchers. SAGE Publications Ltd, 2009. 123

[119] Kjeld Schmidt and Liam J. Bannon. Taking CSCW Seriously: Supporting articulation work. Computer Supported Cooperative Work (CSCW): An International Journal, 1(1- 2):7–40, 1992.

[120] Kjeld Schmidt and Carla Simone. Coordination Mechanisms: Towards a Conceptual Foun- dation of CSCW Systems Design. Computer Supported Cooperative Work, 5(2/3):155–200, 1996.

[121] Phoebe Sengers and William Gaver. Staying open to interpretation: Engaging multiple meanings in design and evaluation. Proceedings of the Conference on Designing Interactive Systems: Processes, Practices, Methods, and Techniques, DIS, 2006:99–108, 2006.

[122] Andrew K Shenton. Strategies for ensuring trustworthiness in qualitative research projects. Education for Information, 22:63–75, 2004.

[123] Jaime Snyder. Visual Representation of Information as Communicative Practice. Journal of the American Society for Information Science and Technology, 66(November):2233–2247, 2014.

[124] Jaime Snyder, Eric P.S. Baumer, Stephen Voida, Phil Adams, Megan Halpern, Tanzeem Choudhury, and Geri Gay. Making things visible: Opportunities and tensions in visual approaches for design research and practice. Human-Computer Interaction, 29(5-6):451– 486, 2014.

[125] Robert Soden and Leysia Palen. From crowdsourced mapping to community mapping: The post-earthquake work of OpenStreetMap Haiti. In COOP 2014: Proceedings of the 11th International Conference on the Design of Cooperative Systems, pages 311–326, Cham, Switzerland, 2014. Springer.

[126] Robert Soden and Leysia Palen. Informating Crisis: Expanding Critical Perspectives in Crisis Informatics. Proceedings of the ACM on Human-Computer Interaction - CSCW, 162:1–22, 2018.

[127] Pitirim A Sorokin and Robert K Merton. Social Time: A Methodological and Functional Analysis. Source: American Journal of Sociology, 42(5):615–629, 1937.

[128] James Spradley. Asking descriptive questions. In The Ethnographic Interview, pages 78–91. Wadsworth, 1979.

[129] Lise Ann St. Denis, Amanda Lee Hughes, and Leysia Palen. Trial by Fire: The Deployment of Trusted Digital Volunteers in the 2011 Shadow Lake Fire. In L. Rothkrantz, J. Ristvej, and Z. Franco, editors, Proceedings of the 9th International ISCRAM Conference, pages 1–10, Vancouver, BC, Canada, 2012.

[130] Standby Task Force. Our activation criteria.

[131] Kate Starbird and Leysia Palen. ”Voluntweeters”: Self-Organizing by Digital Volunteers in Times of Crisis. In Proceedings of the 2011 Annual Conference on Human factors in Computing Systems (CHI ’11), pages 1071–1080, 2011.

[132] Kate Starbird and Leysia Palen. Working & Sustaining the Virtual ’Disaster Desk’. In CSCW ’13, pages 491–502, 2013. 124

[133] Gunnar Stevens and Volkmar Pipek. Making Use: Understanding, Studying, and Support- ing Appropriation. In Socio-Informatics. Oxford University Press, 2018.

[134] Lucy Suchman. Human-machine reconfigurations: Plans and situated actions. Cambridge University Press, 2007.

[135] Andrea H Tapia and Kathleen Moore. Good Enough is Good Enough - Overcoming Dis- aster Response Organizations Slow Social Media Data Adoption. Computer Supported Cooperative Work, 23:483–512, 2014.

[136] John C. Thomas, Yue Pan, Thomas Erickson, Eli Blevis, Catherine Letondal, and Aur´elien Tabard. Avec le temps! Time, Tempo, and Turns in Human-Computer Interaction. In Proceedings of CHI 2013 EA, pages 3303–3306, Paris, France, 2013. ACM.

[137] Anna Tsing. Friction: An ethnography of global connection. Princeton University Press, 2011.

[138] Lea Tufford and Peter Newman. Bracketing in qualitative research. Qualitative Social Work, 11(1):80–96, 2010.

[139] UNHCR. Figures at a Glance, 2019.

[140] Amy Voida, Ellie Harmon, and Ban Al-ani. Homebrew Databases: Complexities of Ev- eryday Information Management in Nonprofit Organizations. In Proceedings of the ACM SIGCHI Conference on Human Factors in Computing Systems (CHI 2011), pages 915–924, 2011.

[141] Amy Voida, Elizabeth D. Mynatt, Thomas Erickson, and Wendy A. Kellogg. Interviewing over instant messaging. In CHI’04 Extended Abstracts on Human Factors in Computing Systems, pages 1344–1347, 2004.

[142] Judy Wajcman. Life in the fast lane? Towards a sociology of technology and time. British Journal of Sociology, 59(1):59–77, 2008.

[143] Mary J. Waller and Sjir Uitdewilligen. Talking to the Room: Collective sensemaking during crisis situations. In Robert A. Roe, Mary J. Waller, and Stewart R. Clegg, editors, Time in Organizational Research, chapter 10, pages 186–203. Routledge, 2009.

[144] Karl E. Weick. Enacted Sensemaking in Crisis Situations. Journal of Management Studies, 25(July):305–317, 1988.

[145] Karl E. Weick. The Collapse of Sensemaking in Organizations : The Mann Gulch Disaster. Administrative Science Quarterly, 38(4):628–652, 1993.

[146] Karl E. Weick and Karlene H. Roberts. Collective Mind in Organizations: Heedful Interre- lating on Flight Decks. Administrative Science Quarterly, 38(3):357–381, 1993.

[147] Etienne Wenger. Communities of Practice: Learning as a social system. Systems Thinker, 9(5):2–3, 1988.

[148] Elaine K Yakura. Charting Time : Timelines as Temporal Boundary Objects. The Academy of Management Journal, 45(5):956–970, 2002. 125

[149] Eviatar Zerubavel. Hidden Rhythms: Schedules and calendars in social life. University of California Press, 1981.

[150] Eviatar Zerubavel. The Standardization of Time : A Sociohistorical Perspective. American Journal of Sociology, 88(1):1–23, 1982.

[151] Eviatar Zerubavel. The Language of Time: Toward a Semiotics of Temporality. Sociological Quarterly, 28(3):343–356, 1987.

[152] Amy X. Zhang and Justin Cranshaw. Making sense of group chat through collaborative tagging and summarization. Proceedings of the ACM on Human-Computer Interaction, 2(CSCW), 2018. Appendix A

SBTF vPD workshop protocol

Workshop logistics and activities:

Since the SBTF volunteers are globally distributed, the workshops will be conducted over pri- vate Slack chat rooms established for the study and not affiliated with SBTF. Participants will be randomly assigned to regional teams according to adjacent time zones (within +/− 6 hours).

Regional teams with more than 20 participants will be divided into smaller sub-teams of 4-5 peo- ple to ensure maximum engagement in activities.

I anticipate organizing 3 regional workshop teams: Asia/Oceania, Middle East/Europe/Africa, and North/South America. Eligible volunteers will be invited to participate in two 90-minute workshops—one for brainstorming and the other for designing. Workshops will be scheduled via online polls at a mutually agreeable time for the team within a 5-day window of time in order to maintain momentum and group cohesion.

Each workshop will consist of warmup and cool down exercises to enable participants to prac- tice skills with the assorted ICTs used in the study and to build team cohesiveness to maximize design outcomes.

Table A.1: Activities and Scripts

Workshop 1: Team design values and roadblock ideation - 90 minutes Activity / Time Description Pre-workshop instructions 127

Pre-W 1 Team formation, Slack invites, workshop scheduling, link to Design Justice zine to participants; pin to Slack channel EMAIL/SLACK Hi everyone. I’m so excited to work with you on this study. A couple of quick housekeeping items and a very fast reading as- signment before we get started on the SBTF study.

Welcome to our Slack channel Please check your email inbox for an invitation to the SBTF Study Slack channel. You should be subscribed to 3 channels: general, random, and your private team.

Meet your teammates Check your private team Slack channel for a welcome message and introduction to your workshop teammates and the research team who will be guiding you during the workshops.

Schedule your team workshops In your private team Slack channel, please work with your team- mates to schedule the two 90-minute workshops. Daytime, night- time, or weekends – whatever is most convenient for the team. It would be best to schedule both of the workshops as close to- gether as possible and to try to wrap up no later than 16 Feb, please. Then, drop me a DM in Slack to let me know what days and times the team has chosen for the two workshops. This free website works well for group scheduling, but you can use any method you prefer. urlhttp://whenisgood.net/

Play with our team whiteboard Please click the following link to accept an invitation to MU- RAL.co, a free collaborative whiteboard. Feel free to draw and play around in the practice mural to get the hang of it. We’ll be using the whiteboard a lot during the workshops so you can share ideas with your team. links vary by team

Pre-workshop reading assignment Before the first workshop, please read this 1-page website about “design justice” linked below and pinned to your team channel in Slack. This is a way of thinking about technology and values that we will use to develop our own guiding principles for the workshops. https://designjustice.org/read-the-principles

Please feel free to reach out to me via Slack DM if you have any questions about the workshops.

Talk soon. Cheers,

Wendy 128

Workshop 1: Team brainstorming and design value decisions - 85 minutes Activity 1A Outcome: Participants understand the purpose of the (5 minutes) study and consent to participation.

Activity: Welcome and introductions of the research team and participants. Reiteration of study purpose, objectives, and consent conditions.

The team convenes in a private Slack team channel.

Teams may also use archival probes for recall (e.g., SBTF data artifacts, an activation Slack channel ex- cerpt, social media examples, and ArcGIS map). These probes will be pinned to the team channels for refer- ence, as needed).

Sample documents to pin to channel:

Google Sheet Copy of 1709 Hospital Status Puerto Rico last version

Slack screenshot Hurricane Maria screenshot.png

Twitter example period tweet.png

Facebook example period FB.png

Tweetdeck example Tweetdeck.png

ArcGIS example SBTF PRmap.png 129

1A SCRIPT PRE-WORKSHOP SETUP ACTIVITY:

Hi all. We are super excited to have your team join us for the SBTF study workshop on MM DATE at TT:TT TZ

Please be sure to have both Slack and MURAL open when you join in.

Team’s LINK to MURAL

Also, it would be helpful to turn on Slack notifications so you don’t miss anything and are alerted to any @here or DM chimes.

Cheers,

Wendy and the research team (Steve, Jordan, Lanea, and Priya)

—————————————-

WORKSHOP DAY SLACK MESSAGE:

A couple of friendly reminders before we get started:

Please login to your team’s MURAL whiteboard. Team LINK Please turn on Slack notifications so you are alerted to any @here or DM chimes

—————————————-

WORKSHOP WELCOME:

:blob-wave: Welcome everyone! I’m so glad you can join us to brainstorm and collectively design new work practices and/or technology improvements that SBTF could potentially use for a future activation.

Our goals today are to: 1. Develop a set of shared design princi- ples to help guide your team. These principles can be helpful for deciding what problem(s) to focus on, what we value as a team, how we work together, and what’s most important for setting priorities.

2. Begin brainstorming new ideas about SBTF work practices and/or the technologies we use during activations. I’ve pinned some previous activation documents in the channel (Google Sheet, social media, map, and Slack examples, etc) that you are welcome to refer back to if that would help. 130

1A SCRIPT 3. Have some “hard fun!” The work we do is so important in- (cont’d) the-moment of a crisis but we often have little time to think about the future. The workshop might not be easy but I hope it will be lively and fun.

Everybody ready? Activity 1B Outcome: Warmup exercise designed to build inter- (10 minutes) team camaraderie.

Each team will chat in their designated Slack private channel with the goal of talking with one another to dis- cover 2 things the team members share in common. 1B SCRIPT SETUP ACTIVITY: Our first activity is a fun ice-breaker we can do right here in our team channel.

Let’s get to know one another a little better. Can we find 2 things that everyone on the team shares in common? Feel free to type out a hobby, a favorite food, a place you like to go on holiday, etc. How much common ground can we find?

Let’s see what we can come up with in the next 5 minutes or so. Ready! Set! Go!

TEAM REFLECTION: What did you find? How many common items did you find? What was your approach to finding out information about each other? Did you find anything surprising? What was it? Activity 1C Outcome: Team develops a set of shared design val- (20 minutes) ues to guide ideation and prototyping activities to em- power participants to claim ownership of the PD pro- cess and mitigate power dynamics between the teams and researcher. Team members have a sense of agency – feel like a stakeholder in the process, are empowered to claim ownership of the process, mitigate potential power dynamics between the group and researchers, and pro- vide values guidance for activity facilitation in their own terms.

This activity focuses on having the group identify and commit to a shared set of values inspired from design justice principles [31] that will guide the par- ticipatory design process. The design justice docu- ment will be pinned to the Slack channel for reference. https://designjustice.org/read-the-principles 131

1C SCRIPT SETUP ACTIVITY: In this activity, we’re going to develop a set of shared design principles to help guide your team in brainstorming and collec- tively designing new work practices and/or technology improve- ments that SBTF could potentially use for a future activation.

These principles can be helpful for deciding what problem(s) to focus on, what we value as a team, how we work together, and what’s most important for setting priorities.

Let’s take a look at the 10 Design Justice principles to give us some ideas about writing down some of our own values that will guide the team’s design decisions and priorities: https://designjustice.org/read-the-principles

—————————————-

TEAM REFLECTION: Let me know when you’re ready to start.

There are several mentions of “community” in the principles. For our purposes as SBTF volunteers, what does “community” look like to you?

Are there any principles that really stand out to you as poten- tially very important values that you would like to help guide the team’s design work?

Are there any principles that feel less relevant to the team in making design decisions about SBTF work practices or technol- ogy improvements? IF STUCK: Maybe ideas or values that don’t fit our community?

Are there any of the principles here that you’re wondering about? That maybe aren’t clear?

Are there any missing values that you think are important for including in the team’s design decisions? IF STUCK: Maybe ideas or values that are more specific to our community than others?

Are there any principles from this list that you’d like to use without modification? We can totally copy and paste from it. :)

Are there any principles from this list that you’d like to borrow and elaborate a bit more to create our own set of values? 132

Activity 1D Outcome: Create team camaraderie and build facility (15 minutes) with using the online whiteboard to contribute collabo- rative sketching.

Cool down exercise: We will present a global map image on the Mural whiteboard. Each participant will be instructed to place a label with their name and current hometown on a map. Then, participants will place a second label on a location they’d like to visit in the future. Return to Slack.

On the main Slack channel, the PI will ask a series of open-ended questions to facilitate discussion between the participants. Finally, the team will begin to explore temporal discord in SBTF activation work practices, technologies, and/or data artifacts.

TEMPLATE: https://app.mural.co/ 1D SCRIPT SETUP ACTIVITY: Wow! You all are amazing! I will put together a document with your team values and pin it to the channel to help guide the brainstorming and design activities.

Okay, let’s switch gears for a minute and jump over to the MU- RAL whiteboard. Here’s the link if you don’t yet have it open: MURAL TEAM LINK-WHERE IN THE WORLD

This activity is called “Where in the World” and we’re going to plot ourselves on a world map. You can navigate the MURAL by clicking on the outline boxes on the right side of the canvas.

MURAL outline button.png

Check out the instructions I have posted on the whiteboard and have some fun!

When you’re done, come back to Slack.

—————————————-

TEAM REFLECTION: Where is a place that you have not heard of before?

What is something that surprised you?

What did you learn about your team?

—————————————- 133

1D SCRIPT Okay, so let’s extend this a bit. (cont’d) Thinking about how the SBTF community is global, let’s take a moment to talk about how different time zones impact the way we work remotely during an SBTF activation. While our team is all in the same or closeby timezones that’s not the case when we’re on an activation. One obvious difference is that most of us are working in a differ- ent time zone than the crisis.

Can you think of other ways that time might be something to keep an eye on?

How does time affect coordinating across the world?

How does time get presented in the technology that we use? (take a peek at the Google Sheets or social media pinned in the channel, if you like)

In what other ways might time be an issue for us in our work or in the technology we use?

IF STUCK: Thinking back to a previous activation, how do you make time for SBTF in between work, family, friends, and other obligations? Tell me a bit about that experience?

IF STUCK: For example, how do you go about making sense of a social media post when the text does not match up with the timestamp? Person says there is “2 hours of diesel left” at the hospital. How do you figure out what that means? When are those 2 hours are up? Like in this Facebook post: period FB.png LINK 134

Activity 1E Outcomes: Teams describe previous activation experi- (30 minutes) ences and gain basic understanding about roadblocks in digital humanitarian work. Teams identify specific in- stances of discord/disruption during activations due to workflows and/or appropriated technologies. Teams can both identify and describe how time impacts their work. Teams identify areas of discord where time/temporality is a factor in SBTF work practices, technologies, and ar- tifacts.

Brainstorming activity: In Slack, the team will be prompted to recall/share a previous activation. Par- ticipants will be prompted to identify any roadblocks or frustrations they have experienced. In a MURAL brainstorming template, the team will be reminded of their design values (activity 1C) and receive a “magic wand” prompt to suspend judgement on the viability of ideas. Participants will notate any experienced acti- vation roadblocks (captured from the earlier Slack di- alog), brainstorm possible improvements to workflows and technologies, and indicate similar ideas on a shared team canvas:

SETUP ACTIVITY: Okay, now we’re going to dive into our first design activity. Ready?

First, let’s think back to a recent SBTF activation that stands out in your memory. It’s fine if you all have different activations in mind. No worries.

Can each of you tell me a little about what that activation expe- rience was like?

• Which activation were you involved in?

• What kind of work did you do?: Look for information, build a map, coordinate volunteers, work behind the scenes with the activating partner, etc.?

• Where were you when you helped with the activation?: At work, home, in a coffee shop or someplace else? 135

1E SCRIPT Just pop some thoughts about your activation experience here. (cont’d) Hazy memories are good too!

So cool! I love hearing about how everyone gets involved in help- ing people after a crisis.

Now, let’s think a little about some of the inevitable frustrations we all encounter during an activation when we all want to help but the tech doesn’t do what we need or a usually helpful work practice isn’t helpful for this activation.

What things come to mind when you think about a roadblock you’ve encountered when volunteering on an activation? For ex- ample, sometimes it is hard to know where to even start. Or how to decide what information is timely and reliable. Or how to know who’s working on which task so we’re not duplicating efforts.

I’ve pinned some SBTF documents to the channel if you need a reminder or creative boost about an activation experience.

Let’s hop over to our MURAL to brainstorm some examples of roadblocks, frustrations, or just a sense of discord that can hap- pen during an activation.

Navigate the MURAL by clicking on the outline boxes on the right side of the canvas in number order.

If you need to move the canvas, use this hand button. If you want to zoom in/our, use the slider.

Follow the directions on each activity, beginning with 1. Get- ting Started. If you have any questions, hop back over here on Slack and I will try to assist.

Remember, don’t worry about the viability of an idea. Unleash your creative superpowers! :hero:

When you’re done with activity 2. Activation Roadblocks come back to Slack to share a few of your ideas. 136

1E SCRIPT TEAM REFLECTION: (cont’d) These are such great ideas! Let’s talk about the activation road- blocks first.

Can somebody fill me in on Issue 1?

How about Issue 2?

Help me understand Issue 3?

Issue 4 is interesting. What’s going on there?

Let’s talk about Issue 5 now.

Last, tell me more about Issue 6.

Great! Now, let’s try brainstorming some solutions to these roadblocks.

Feel free to brainstorm ideas that would solve any of the roadblocks that were identified by your team. Again, wild ideas are welcome! Remember to suspend your judgement and think BIG!

You’re welcome to talk amongst yourselves here if you’d like to collaborate or ask questions about a roadblock.

When you’re done with this activity, come back to Slack to share a few of your ideas. :sunglasses:

Head back over to MURAL and begin activity 3. Heads down.

—————————————-

Hello again. Can’t wait to chat about some of your brainstorms. Please choose 2-3 of ideas that you generated from your list that you’re excited about.

@NAME1 what’s your best couple of solution ideas?

@NAME2 tell me some of your ideas

@NAME3 can you share your brainstorms?

@NAME4 tell us what you’ve got

@NAME5 what did you come up with?

These ideas are amazing! It gives us a great place to start in the second, final workshop. Thank you. 137

Activity 1F Outcome: Team reflects on workshop experience to (5minutes) build collegiality within team and sense of accomplish- ment. Reflection and feedback provide further analyti- cal data.

Team debrief in Slack to conclude workshop 1 to share thoughts, recalibrate activities from feedback, etc. A link to an anonymous survey will also be distributed to the participants to solicit procedural feedback.

Question prompts will include: “What did you take away from today?” and “What would you like to ask about?” as a way to ask questions, share concerns, re- quest clarifications, etc. Participants will then be di- rected to a Google Form to anonymously share their feedback about each of the day’s activities via a 5-point Likert scale. The Google Form also offers a free text section for optional written comments and feedback. 1F SCRIPT SETUP ACTIVITY: That’s it! We’ve come to the end of Work- shop 1.

Is there anything you’d like to ask about in terms of next steps or would you like to share any thoughts on take-aways from to- day’s workshop?

—————————————-

In the final workshop, we will have an opportunity to sketch some of our ideas. I’ll post a link to a practice MURAL here if you’d like to play around with the drawing tools ahead of time.

—————————————- Here’s a link to an anonymous survey where you can provide some feedback about the activi- ties we did today. Your name will not be recorded. You’re also welcome to leave comments about your experience.

Thank you so much! Looking forward to seeing what you design in Workshop 2!

https://forms.gle/fcFhUDjg4XNgcgnV8

Workshop 2: Team co-design and prototyping priorities - 90 minutes Activity / Time Description Pre-workshop instructions 138

Pre-W2 Team formation, Slack invites, workshop scheduling, link to Design Justice zine to participants; pin to Slack channel EMAIL/SLACK CONFIRMATION EMAIL:

Hi all. We are super excited to have your team join us again for SBTF study workshop2 on MM DATE at TT:TT TZ

Please be sure to have both Slack and MURAL open when you join in. Team’s LINK to MURAL

Also, it would be helpful to turn on Slack notifications so you don’t miss anything and are alerted to any @here or DM chimes.

See you soon!

Wendy and the research team (Steve, Jordan, Lanea, and Priya)

—————————————-

WORKSHOP DAY SLACK REMINDER MESSAGE:

A couple of friendly reminders before we get started today at TT:TT TZ

Please login to your team’s MURAL whiteboard. Team LINK

Please turn on Slack notifications so you are alerted to any @here or DM chimes

See you on Slack in a bit! Activity 2A Outcome: Re-center team on study purpose, objectives, (5 minutes) and consent conditions.

Welcome back team. Reflect on Workshop 1 accom- plishments. Opportunity for questions and clarification. 2A SCRIPT SETUP ACTIVITY:

:blob-wave: Welcome back! Does everyone have MURAL open and their Slack notifications turned on? :confused-parrot:

I’m so glad you can join us for this last session to brainstorm and collectively design new work practices and/or technology improvements that SBTF could potentially use for a future acti- vation. 139

2A SCRIPT Our goals in our final workshop are to: 1. Revisit the design jus- (cont’d) tice principles the team developed during the first workshop to help refocus our attention on values, needs, and priorities for our design work today.

2. Review the lists of roadblocks and brainstorms for SBTF work practice and/or technology improvements from the first workshop in order to select a few to focus on today.

3. Sketch out some initial ideas on the shared whiteboard that the team thinks would help to improve SBTF work practices and/or the technologies we use during activations.

4. Let’s have some fun! The work we do is so important in-the- moment of a crisis but we often have little time to think about the future. The workshop might not be easy but I hope it will be lively and fun.

During this workshop, feel free to use some of the pinned activa- tion documents in the channel (Google Sheet, social media, map, and Slack examples, etc). I can paste any of the images into the MURAL canvas if you would like to sketch on them. Just let me know.

Okay, everybody ready? Do you have any questions before we jump in?

TEAM REFLECTION:

... Activity 2B (5 Outcome: Warmup exercise to prompt creative thinking minutes) for workshop activities.

This brief warmup exercise is designed to prompt cre- ative thinking. Participants will each draw a vase on a single MURAL whiteboard sectioned-off for each team. Then, each team member will draw a different way for people to enjoy flowers in their home. Convene all the participants on the main Slack channel to compare the two different drawings on the whiteboard and discuss how reframing a question changed the proposed solu- tion. 140

2B SCRIPT SETUP ACTIVITY: Okay, let’s start off with something fun :art:

In the MURAL whiteboard, I’ve set up a canvas with a section for each of you. This activity is called “How to enjoy flowers.”

Feel free to use any of the MURAL drawing tools on the left toolbar.

Use the outline on the right side of the canvas to navigate to the area with your name on it.

On the left side of the box, draw a vase to hold a bouquet of flowers in your home.

On the right side of the box, draw a quirky, unconventional way to display the bouquet that would make the flowers last twice as long and that is not a vase. Get as wild and creative as you want!

The trick here: we have just 2 minutes to draw 2 sketches.

When you’ve drawn two quick sketches, come back to Slack.

Ready! Set! Go!

MURAL TEAM LINK-HOW TO ENJOY FLOWERS

TEAM REFLECTION:

Your drawings are so great! Tell me, how did the time factor (to make the flowers last twice as long) play into your thinking about a quirky way of displaying the flowers? Activity 2C (3 Outcome: Reflections on Design Justice principles from Work- minutes) shop 1 discussion

Team reviews a poster of their selected design justice principles to more deliberately incorporate these values into continued brainstorming on roadblocks and solu- tions.

A poster representing the team’s design justice princi- ples will be pinned to the Slack channel and posted in the Mural canvas to guide subsequent activities. 141

2C SCRIPT Before we jump in, let’s review the design justice principles that the team decided on last time we met. At each stage of the de- sign work today, we will want to keep these principles in mind to reflect on our ideas to make sure the designs are upholding the team’s values.

I have pinned this poster in the team Slack channel and will also post it to the MURAL canvas. Sound good? Activity 2D (8 Outcome: Convergence on roadblocks identified during minutes) Workshop 1.

Team reviews initial roadblock brainstorms, dis- cuss/debate ideas on Slack. 2D SCRIPT SET UP ACTIVITY:

Let’s take a moment to quickly review the roadblocks that the team identified in the last workshop to refresh our memories.

If you recall, we brainstormed some of the frustrations that the team has encountered with workflows and/or technologies during an activation.

Upload team’s roadblock canvas image to Slack

TEAM REFLECTION:

These are fabulous ideas to design for!

Are there any questions you’d like to ask of your team to clarify a sticky note before we begin to sort them?

Do you see any obvious themes emerging from the sticky notes? What stands out to you? Activity 2E (10 Outcome: Thematic convergence on roadblocks identi- minutes) fied in Workshop 1

Team sorts roadblocks into thematic groupings on MU- RAL.

Introduce MURAL in-app chat function to facilitate team discussion/debate while grouping roadblock sticky notes.

Team returns to Slack to discuss overarching themes and to compare the themes against the team’s design justice values. 142

2E SCRIPT SET UP ACTIVITY:

Okay, so now we need to start sorting the roadblock sticky notes into groups to help us narrow down our design focus.

We’ll do this in MURAL.

Use the outline on the right side of the canvas to navigate to the section labeled 2. Roadblock themes and follow the instructions for dragging the roadblock sticky notes into common themes.

But before we hop over there ...

Let me show you where you can find the chat function on the canvas that will allow you to talk with one another during this activity. This one will require some coordination between every- one.

Chat button

We’ll take about 5 minutes to do this and I’ll set a timer on the canvas for you. When you’re done, come back to Slack.

Does that make sense? Are we ready to go?

TEAM LINK

TEAM REFLECTION: Can you tell me about the different themes you came up with?

Is everyone in agreement about the themes or would you like to make some changes now that we’ve talked about them?

—————————————-

Now let’s take a look at the team’s design justice principles.

For this purpose, let’s assume that the “designer” is someone at Google, Slack, Twitter or Facebook who is responsible for the technology that SBTF uses and is open to hearing about com- munity needs.

Do you find that any of the principles relate to the themes? In what way or, even, why not? Activity 2F (5 Outcome: Set priority on roadblock themes. minutes) Team determines the importance and feasibility of road- block themes. 143

2F SCRIPT SET UP ACTIVITY:

This is so great and will really help us move toward well- informed designs.

But we do need to set some priorities in order to collectively de- sign new work practices and/or technology improvements that SBTF could potentially use for a future activation.

Now’s the time to make some tough choices.

We need to evaluate each of the roadblock themes in two ways: Importance and Feasibility.

How might we think about the “importance” of a theme? Any ideas?

—————————————-

We are only responsible for imagining the workflow or technol- ogy improvement.

We will forward the idea to the experts to make. So don’t worry too much about your own skill level to make the design work.

With that in mind, how might we think about defining “feasibil- ity”?

—————————————- Great!

In section 3. Prioritize on the MURAL canvas . . . that I will re- veal in just a sec ... follow the instructions to make a new sticky note for the theme label and place it on the priority grid.

So it’s not total chaos :joy: . . .

I assigned each of you to a theme (or two) to set the initial pri- ority. Later, we will make a team decision about the priorities in order to decide which theme roadblocks move forward to the design phase.

@NAME: Theme 1 @NAME: Theme 2 @NAME: Theme 3 @NAME: Theme 4 @NAME: Theme 5

Feel free to use the MURAL chatbox to talk with each other.

When you’re done, come back to Slack. Ready?

TEAM LINK

TEAM REFLECTION: 144

2F SCRIPT You all did an amazing job! (cont’d) I’m curious how you decided whether something was important and/or feasible?

Any thoughts on where the themes appear on the grid?

IF STUCK: How is team setting priorities? (e.g. working re- lations, supporting communities needs, efficiency, productivity, resilience of those they are providing the data for?) Activity 2G Outcome: Team votes on theme priorities (importance and feasibility) 2G SCRIPT Now let’s vote on the grid placements.

We need to be swift!

Everyone has 3 votes that they can cast with the stars above the grid.

To vote, drag each of your assigned stars and place it on any of three theme sticky notes that you want to advance to the design round.

The MURAL chat box is available to discuss/debate.

When you’re done voting come back to Slack to talk about it.

Go vote!

TEAM LINK

TEAM REFLECTION:

How did it go?

For the themes that you starred, do you also agree with the im- portance and feasibility placement? Or would you like to debate it?

Are there any themes that didn’t earn enough votes to move for- ward that anyone would like to advocate for?

Are we in agreement or should we talk this out a bit more? Activity 2H (5 Outcome: Articulate a problem statement minutes) 145

2H SCRIPT Last task before we start designing!

Now that we have focused our attention to a small set of road- blocks, grouped them within a shared theme, and ranked them on the basis of importance and feasibility, we have one last task to take care of.

We need to briefly state for the designer what problem the team is trying to solve. We’ll do this here in Slack.

Can we frame the top priority theme as a “How might we. . . ” statement?

For example, if the theme is “too much information” . . . a prob- lem statement could be “How might we help SBTF volunteers to overcome information overload?”

Any questions? Ready to try to write your own problem state- ment?

TEAM REFLECTION: Activity 2I (15 Outcome: Imagine and sketch designs to address the minutes) problem statement

The team uses the Slack channel and MURAL white- board to discuss, sketch, improvise, re-prioritize, envi- sion, and prototype design mockups of the priority (im- portant/feasible) technology improvements 2I SCRIPT Let’s start designing!

We have a problem statement: TEAM RESPONSE

A theme: TEAM RESPONSE

Roadblocks for inspiration: TEAM RESPONSE

Imagine as many possible solutions to the roadblocks as possi- ble. Use any of the tools on the left menu bar of the canvas to convey your idea, e.g., sticky notes, drawings, icons/clipart, or download any of the files pinned to this channel.

Remember, don’t worry about the viability of an idea. Unleash your creative superpowers! :hero:

Feel free to use the in-canvas chat tool to talk and collaborate with one another. You are free to work alone or with others and play off each other. If you like an idea drop a star on it. 146

2I SCRIPT If you need help, ping me in the MURAL chat or here in Slack (cont’d) and I will try to assist.

Sound good? Any questions?

—————————————-

Let’s hop into our Really Big Sketch Pad in MURAL. You have about 15 minutes to brainstorm, sketch, and dream.

TEAM LINK

—————————————-

In MURAL chat:

Can we take a minute or two to quickly review your team’s de- sign ideas? Feel free to drop a star on an idea you really like. Activity 2J (15 Outcome: Identify, reflect, and describe potential work minutes) practice and/or technology improvements.

Team reconvenes on Slack to present and prioritize whiteboard sketches to the research team.

The participants reconvene in Slack to review, discuss, and negotiate about the priorities. Here, they will fur- ther refine which ideas to focus their attention during this workshop by moving the Mural sticky notes into the center of the whiteboard to signify consensus about the focus of the design workshop. SCRIPT 2J SETUP ACTIVITY:

Wow! Such great ideas!

Let’s take the next few minutes to share your thinking behind some of the sketches and ask questions of each other.

Then, we’ll vote on a few to move forward to our designer.

Any initial thoughts?

TEAM REFLECTION: 147

Activity 2K Outcome: Discussion of research integration of the team’s final design mockups and next steps.

The PI will discuss with all participants on the main Slack channel how their work, design decisions, prior- ities, mockups, etc. will be reflected in the research project. The research team will also initiate a conver- sation with the teams on the next steps. 2K SCRIPT SETUP ACTIVITY:

This was so fun and the research team is very grateful to you for sharing your time and expertise with us. SBTF is THE best!

The plan moving forward is to collect all of the ideas created by the different teams and begin prototyping the most promising ones. The first round of design work will begin with those that won top votes.

Just as a reminder: in the dissertation and research papers that I will be writing about the workshops, your names will not be used and your identities (Slack profile pic, etc) will be masked.

Do you have any questions for me? We can chat here or you can ping me via DM.

TEAM REFLECTION: Activity 2L (5 Outcome: Team reflects on workshop experience to minutes) build collegiality within team and sense of accomplish- ment. Reflection and feedback provide further analyti- cal data.

Team debrief in Slack to conclude workshop 1 to share thoughts, recalibrate activities from feedback, etc. A link to an anonymous survey will also be distributed to the participants to solicit procedural feedback.

Question prompts will include: “What did you take away from today?” and “What would you like to ask about?” as a way to ask questions, share concerns, re- quest clarifications, etc. Participants will then be di- rected to a Google Form to anonymously share their feedback about each of the day’s activities via a 5-point Likert scale. The Google Form also offers a free text section for optional written comments and feedback. 148

2L SCRIPT SETUP ACTIVITY:

That’s it! We’ve come to the end of Workshop 2.

Is there anything you’d like to ask about in terms of next steps or would you like to share any thoughts on take-aways from to- day’s workshop?

—————————————-

Here’s a link to an anonymous survey where you can provide some feedback about the activities we did today. Your name will not be recorded. You’re also welcome to leave comments about your experience.

Thank you so much!

https://forms.gle/ePhkKAq8rmaBFw4d9