See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328689561

Personal Informatics in Interpersonal Contexts: Towards the

Design of Technology that Supports the Social Ecologies of
Long-Term Mental Health Management

Article in Proceedings of the ACM on Human-Computer Interaction · November 2018

DOI: 10.1145/3274396

CITATIONS READS
136 2,131

5 authors, including:

Elizabeth Murnane Beck Tench

Dartmouth College University of Washington Seattle
79 PUBLICATIONS 1,894 CITATIONS 2 PUBLICATIONS 136 CITATIONS

SEE PROFILE SEE PROFILE

Jaime Snyder
University of Washington Seattle
56 PUBLICATIONS 700 CITATIONS

SEE PROFILE

All content following this page was uploaded by Elizabeth Murnane on 10 November 2018.

The user has requested enhancement of the downloaded file.

Personal Informatics in Interpersonal Contexts:
Towards the Design of Technology that Supports the
Social Ecologies of Long-Term Mental Health Management

ELIZABETH L. MURNANE, Stanford University, USA

TARA G. WALKER, University of Colorado Boulder, USA
BECK TENCH, University of Washington, USA
STEPHEN VOIDA, University of Colorado Boulder, USA
JAIME SNYDER, University of Washington, USA
Personal informatics systems for supporting health largely grew out of a “self”-centric orientation: self-tracking,
self-reflection, self-knowledge, self-experimentation, self-improvement. Health management, however, even
when self-driven, is inherently social and depends on a person’s direct relationships and broader sociocultural
contexts, as an emerging line of research is coming to recognize, study, and support. This is particularly true
in the case of mental health. In this paper, we engage with individuals managing the serious mental illness
bipolar disorder and members of their support circles to (a) identify key social relations and the roles they play
in condition management, (b) characterize patients’ complex interactions with these relations (e.g., positive
or negative, direct or peripheral, steady or unstable), and (c) understand how personal informatics mediates
these recovery relations. Based on these insights, we offer a model of this social ecology, along with design
implications for personal informatics systems that are sensitive to these interpersonal contexts.
CCS Concepts: • Human-centered computing → Human computer interaction (HCI); Collaborative
and social computing;
Keywords: Personal Informatics; Quantified Self; Collective Intelligence; Collaborative Sensemaking; Self-
Management; Mental Health; Serious Mental Illness; Bipolar Disorder
ACM Reference Format:
Elizabeth L. Murnane, Tara G. Walker, Beck Tench, Stephen Voida, and Jaime Snyder. 2018. Personal Informatics
in Interpersonal Contexts: Towards the Design of Technology that Supports the Social Ecologies of Long-
Term Mental Health Management. In Proceedings of the ACM on Human-Computer Interaction, Vol. 2, CSCW,
Article 127 (November 2018). ACM, New York, NY. 27 pages. https://doi.org/10.1145/3274396

1 INTRODUCTION
Recent years have seen a significant increase in the availability of technologies for tracking various
details of one’s life. The term Personal Informatics (PI) was coined to refer to such tools aimed at
helping users collect and reflect on personal information [57]. While people use PI for a number of
reasons (e.g., to track finances, document visited locations, out of curiosity, or to receive rewards
Authors’ addresses: Elizabeth L. Murnane, Computer Science, Stanford University, Stanford, CA, USA, emurnane@stanford.
edu; Tara G. Walker, Media Research and Practice, University of Colorado Boulder, Boulder, CO, USA, tara.g.walker@
colorado.edu; Beck Tench, Information School, University of Washington, Seattle, WA, USA, [email protected]; Stephen Voida,
Information Science, University of Colorado Boulder, Boulder, CO, USA, [email protected]; Jaime Snyder, Information
School, University of Washington, Seattle, WA, USA, [email protected].

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the 127
full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specific permission and/or a fee. Request permissions from [email protected].
© 2018 Copyright held by the owner/author(s). Publication rights licensed to ACM.
2573-0142/2018/11-ART127 $15.00
https://doi.org/10.1145/3274396

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:2 E. L. Murnane et al.

on social networking sites) [29], most users are interested in capturing and accessing data relevant
to health, specifically in an effort to change behavior (e.g., to lose weight, quit smoking, or improve
sleep) or as part of self-managing a chronic condition [77, 92]. Quantified Self (QS) practices, which
refer to individuals examining various facets of their lives and bodies through self-tracked data,
often similarly have the goal of optimizing health and wellness [62, 79].
Initially, the design of PI systems for health typically emphasized an “ego-centric”perspective
focused on a single person [53], reflecting a tendency to self-orient engagement with personal
data by placing individuals at the center of tracking practices. The term itself, personal informatics,
stresses this model of tracking as an individual activity.
Our research extends a growing body of literature that reflects an increasing interest in studying
and designing for more collective PI practices, such as socially sharing personal data, tracking on
someone’s behalf, and pooling communal data [10, 53, 61, 66, 92, 93]. Further, even when PI activities
do not explicitly involve multiple parties, independent collection and interpretation of data is still
shaped by social institutions [63]. These evolving use cases throw a number of sociotechnical issues
into sharp relief, such as representation, privacy, provenance, brokerage, sensemaking, knowledge
production, and decision making. In this paper, we offer an operationalization of these issues based
on Bronfenbrenner’s ecological model of social relations [14, 15] in order to strengthen and expand
the theoretical tools available to PI researchers seeking to move beyond ego-centric applications.
We use our work with individuals who self-track as a form of mental health therapy in order to
demonstrate how our model can be applied even when social relations are especially complex.
Indeed, the deep embedding of health management practices in social contexts is particularly
salient in the case of serious mental illness (SMI), which affects 1 in 25 adults in the United States
in a given year [80]. Prior work has established that many individuals facing SMI adapt off-the-
shelf PI tools (e.g., self-tracking apps, wearables, spreadsheets, calendars, etc.) for the long-term
management of their condition [51, 69, 76]. However, this comes with challenges such as a lack of
support for tracking SMI-specific indicators, usability issues that impede logging during unwell
periods, and the fact that technology use in general can disrupt emotional balance by interfering
with sleep or fueling a sense of social inadequacy [70].
While these types of personal data do not necessarily make it into standardized healthcare
records [104], they are often informally shared with clinicians and care managers in the course of
establishing lifelong self-care practices. Further, when individuals with SMI share their self-tracking
information with members of their support networks, a diverse and potentially revolving set of
interpersonal connections (referred to in this paper as “social relations” or “stakeholders”) are
activated around personal data practices, influencing and contributing to the use of these traces in
ways that extend beyond single-user requirements.
In working in this space, we have been particularly inspired by the notion of “relational recovery”
[86] from the mental health domain that overtly acknowledges the role that family, friends, antago-
nists, peers, organizations, cultures, and numerous other potential relations play in the process of
coping with SMI. However, even this term — referring to “recovery” — can be problematic because
there is no cure for many SMIs, and managing chronic mental illness often spans a lifetime. Because
of tensions like this, the social implications of siloed tracking and personal data practices become
highly acute in the long term management of SMI. However, as shown by recent work highlighting
the beneficial impacts of social support and advocating for more socially-embedded approaches to
interaction design [2, 39, 91], these issues apply to broader populations as well.
To better understand how PI systems can leverage a model of social and technological relations,
we explore the connections and bonds — both positive and negative, weak and strong, current and
past — between individuals with SMI and the social relations that surround them. Through a series

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:3

of one-on-one interviews with individuals with SMI (N =14) as well as focus groups involving both
individuals and stakeholders (N =8), this paper undertakes that conversation.
Our specific contributions are as follows:
• A characterization of the social relations and roles in dynamic support systems surrounding
the self-management of SMI.
• The operationalization of a model grounded in ecological systems theory and our empirical
data that formalizes these social relations and roles.
• A series of implications of this work for PI system designers who want to leverage the
structure of a social relations model to address collective and dynamic contexts of use.
As such, our work contributes to recent PI research in several ways. First, while a number of
studies have focused on patient–provider relationships, including with regard to electronic health
records (EHR) and the sharing of patient-generated data in clinical settings [17, 20, 58, 74] (or
the reluctance to share given issues of perceived relevance, time constraints, or provider rapport
[58]), we focus on the “long tail” of self-management characterized by individuals’ ad-hoc, SMI-
specific, and “vernacular”, or everyday, PI practices [8, 16, 100] that are often performed outside of a
healthcare provider’s oversight. In particular, the longitudinal nature of these experiences means that
social relations will come and go, evolve and disappear, providing a rich set of complications when
applying a social ecology model. Second, while prior work has begun to focus on understanding
social PI practices (e.g., sharing, comparing, and pooling data), we are also interested in the ways that
the introduction of tracking artifacts (e.g., tools, data, visualizations, etc.) influence social practice
in emergent and at times unpredictable ways [91]. Finally, prior conceptualizations of the social
ecosystem that surrounds personal health management portray a relatively static constellation
of stakeholders; in contrast, we highlight the ecological dynamism of lived social experiences for
individuals coping with SMI.

2 RELATED WORK
2.1 Serious Mental Illness — The Case of Bipolar Disorder
In this paper, we focus on individuals with bipolar disorder (BD), an SMI that affects approximately
60 million people globally [78]. BD is characterized by fluctuations in mood as well as disruptions
to biological and behavioral rhythms (particularly physical activity and sleep). BD is chronic and
has no cure; those with the condition must manage it for a lifetime. Recognized as one of the ten
most debilitating illnesses worldwide, 1 in 5 cases results in suicide [3]. Quality of life for those with
BD can be quite poor due to relationship problems, interrupted education or undermined career
advancement, and the physical and emotional stress of indefinitely managing the disorder [36].
These negative impacts are not limited to the individual — they can have destructive and costly
effects on family, friends, and society, with the frequency of episodes having an additive association
with social morbidity, including divorce, injury to oneself or others, and minor crime [60].
Encouragingly, recent research on behavior-based interventions has shown that tracking and
stabilizing everyday activities such as light exposure, sleep–wake routines, and social rhythms
can result in pronounced functional improvements and minimized symptom severity for people
with BD [32, 72]. Clinical guidelines therefore encourage incorporating daily self-monitoring into
condition management [48]. Furthermore, self-tracking can provide a sense of agency for people
whose disorder makes them feel out of control [76] and can help individuals establish and maintain
a sense of identity as their condition evolves [45]. However, the unpredictable mood swings that
characterize BD can be difficult to recognize and self-assess by the person experiencing them,
resulting in warning signs going un-noted and unrecorded [55, 60]. At the same time, appropriate
disclosure and communication about BD experiences can help build stronger and more empathetic

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:4 E. L. Murnane et al.

social support networks [38]. These are all areas in which improved tools for more collective capture,
verification, and interpretation of personal data can positively impact the long-term management
of BD.
Coming to grips with both public and private impacts of periodic, unpredictable fluctuations in
mood and behavior is central to the experience of living with this SMI. Individuals with BD must
navigate complex social contexts where the benefits of support and the risks of stigma regularly
collide. Altogether, BD therefore makes for a particularly compelling case study in considering
collective social engagement in PI — not only due to the opportunities for PI advancements to
improve quality of life for this population, but also because the ways this SMI unfolds over time
brings issues of personal information, complex social contexts, and dynamic user requirements
into specific relief.

2.2 Personal Informatics for Mental Health

PI tools, biometric sensors, and semi-automated techniques originally developed for a more general
set of users (e.g., to track movement, communication, and context [33]) have become therapeutic
resources for individuals managing SMI [76]. As early as 2006, the U.S. Institute of Medicine
recognized the potential of innovations in mobile technology to transform mental health services
by providing more continuous and precise information on patient-specific behavior, symptoms,
and medication effects [81].
However, it can be quite challenging to develop PI technologies that are sensitive to the unique
needs of individuals with mental health conditions like BD (e.g., due to degraded functional abilities
[71], inaccurate reporting [55], or lapsed adherence [44]). Systems like MoodRhythm [68, 99] and
MONARCA [5, 34] are therefore exploring passive data collection and symptom detection that
relieve some of the burdens and limitations of purely manual tracking.
In addition, off-the-shelf tracking technologies can support healthy self-awareness, but there
is a general risk that the ways data is presented back to users can foster hyper self-scrutiny and
unrealistic normative expectations of health or identity for vulnerable individuals like those with
BD [62]. These standardized data representations do not always align with the lived experiences
of individuals, potentially complicating the process of using these personal records to build un-
derstanding and empathy among family and friends [52]. Further, streaming or “always on” data
collection can make it challenging for individuals to practice intentional self-disclosure [69], raising
important issues of privacy and data curation [75]. Finally, efforts to integrate self-tracking data
into EHR systems and broader clinical practices are still in the early stages, which can lead to
individuals with BD experiencing feelings of resentment and a sense of self-tracking futility when
PI data goes unaccepted or unvalued by doctors with whom it is shared [76].

2.3 (Inter)-Personal Health Informatics

In examining the social context for tracking tools used in the long-term management of SMI like
BD, the most visible and studied relationship is that between patient and provider (e.g., therapists,
psychiatrists, general practitioners, neurologists, and so on). In fact, healthcare has long been a
topic of interest to the CSCW community, often focusing on the collaborative nature of healthcare
practices as well as the implications for designing technology that supports such work [31].
In formal care settings such as clinics and hospitals, patient health information (PHI) is docu-
mented and coordinated through medical records [9]. A significant body of literature exists around
the role of PHI as not only information artifacts but as mediators of collaborative healthcare work.
Relatively new innovations, like the EHR and a host of online portals for managing personal health
information, promise to improve efficiency and overall care quality, in part by improving providers’
ability to communicate and share data, even when remotely located [42]. While such medical

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:5

records are provider-facing, personal health record (PHR) systems are patient-facing tools for
collecting, managing, and sharing information (e.g., about medical history and medications) [98].
In the SMI context, a common use case for such a system has been to enable patients to fill in
self-assessment data as part of their daily routine, with a psychiatrist using the system’s clinical
portal to monitor all of his or her patients as well as review historical data together with patients
during therapy sessions [37].
Recently, work has begun to examine how patients and providers can also coordinate over
more vernacular or individualized forms of patient-generated data, captured using commercial
PI applications and other non-clinical logging instruments [20, 46, 74, 104]. However, researchers
have observed barriers to such data being effectively used as part of formalized healthcare practices.
For instance, the volume of information collected can be overwhelming and hard to interpret
[94]. The introduction of digital systems can also degrade face-to-face interactions by drawing
attention away from patient-clinician communication [58]. As self-tracked information becomes an
increasingly significant part of the health data equation, researchers have therefore been making
efforts to develop “shared care" systems that specifically support the integration of meaningful PI
data streams as well as active engagement of both the professional care team and the patient in
monitoring and treatment [6].
Researchers have also investigated how informatics technology can support collaboration around
health management outside of the clinical setting, particularly in the context of family-centered
practices — for instance to track on behalf of other family members including children [50], deliver
digestible representations of detailed family-level data streams, and manage privacy concerns
especially in complex arrangements such as families with divorce [85]. Work looking at coordination
among patients and close caregivers, including spouses, has further identified the importance of
determining and balancing values (e.g., with respect to shifting responsibilities, mutual support,
and patient autonomy) when designing systems for collaborative self-care [10, 82].
The personal informatics literature has also identified a number of other considerations around
if, how, with whom, and why people share health-related PI data. Much of this work has been done
in the context of contemporary sociotechnical systems like social media. These platforms have
been taken up by individuals as a way to record, broadcast, and reflect on their personal health —
for instance, by using Instagram to track and share food data [19], Twitter to seek quick health
advice [64], or Facebook and condition-specific sites to connect with a wide support network of
peers and caregivers [75]. Recent work on mental health specifically has explored the potential for
remote peer support to be delivered through online chat interfaces [83].
Willingness to share personal health data in such venues is by no means the default stance,
however, often due to perceived threats to reputation or privacy [13]. For example, research has
observed that while sharing heart rate data might enhance a sense of connection for couples, such
disclosure is seen as less desirable when the recipient could be a boss or other relation who may
misinterpret the data or pass negative judgment [95]. Other studies suggest that people are more
likely to share summaries, which are more easily digested and privacy-preserving than raw data,
which might be overwhelming or too revealing [26, 27].
Altogether, prior work offers pointers to the presence of socially-driven attitudes, assumptions,
and values influencing personal data practices. While a variety of motives can influence how and
why people share their PI data, such as to receive feedback or emotional support, produce a sense of
accountability, provide information, or manage impression [28], little is known about the ways that
these perceptions and values change over time as a result of social factors [91]. Further, sensitive
conditions like SMI carry a host of vulnerabilities that heighten concerns regarding what data is
visible to whom.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:6 E. L. Murnane et al.

In order to support evolving personal data practices in health and wellness contexts, the PI
research and design community critically needs ways to consider social relations and roles, especially
in terms of how they might change over time. Our work draws on and extends prior efforts by
adapting and applying a model of social and technological relations to the use case of self-tracking
to support SMI. In the process, we identify design implications for supporting collective engagement
in personal health across clinical and non-clinical settings and over time as social support systems
ebb, flow, and evolve.

3 CONCEPTUAL MODELS OF HEALTHCARE STAKEHOLDERS AND ECOLOGIES

In contrast to a “single-loop” approach to personal health informatics that focuses on self-tracking,
self-reflection, and self-improvement chiefly from the perspective of the individual [53], recent work
has advocated for a “double-loop” model involving both the patient and the clinician in tandem [6].
Still necessary, however, is accommodating the broader network of relations, beyond professionals,
potentially involved in an individual’s health management in various implicit and explicit capacities.
Devices used for self-monitoring often lack “situated support” across the full range of care networks,
which need to at a minimum acknowledge not just the needs of the individual but also the needs of
these wider stakeholder connections in the use of self-monitoring health technologies [2, 39].
Partners, relatives, and professional caregivers are commonly identified as stakeholders [25]
who could benefit from shared information spaces. However, research specifically focusing on the
context of SMI care has identified at least eight such stakeholder groups (e.g., patient associations,
financiers, researchers, and other organizational entities that support institutionalized care such as
clinics and facilities, police departments, social work, and home care services) [11]. Still, recent
research advocates for the deeper study of additional physical, organizational, social, and cultural
contexts of care to inform the development of collaborative informatics technologies that capture
and support health-related activities in both clinical and daily-living environments [84]. Through
interviews and focus groups with individuals managing SMI and the people who provide them
direct support, we seek to better understand not just the roles of direct and professional stakehold-
ers involved in mental health care, but any type of social relation implicated in an individual’s
bipolar experience, from close friends and co-workers to peers in community support groups.

Salient to this investigation is Bronfenbrenner’s MACROSYSTEM

Ecological Systems Theory (EST) [14, 15], which de- Attitudes and ideologies
of the culture

scribes a spectrum of macro-level influencers through EXOSYSTEM

micro-level relational support units. The EST model MESOSYSTEM

Industry Neighbors
(see Figure 1) captures the breadth of influences on MICROSYSTEM
an individual’s life, which can operate at various de- Family Peers

grees of proximity, from biological attributes all the INDIVIDUAL

(sex, age,
health, etc.)
way to the broad sociocultural contexts that structure Mass
School Church
Social

human interactions [21]. Research connecting EST to media Health services services

SMI has recognized that mental health issues must be

viewed together with cultural factors, which influence Local politics

the acceptance, stigmatization, and even definition of

mental illness [7, 21, 89, 97]. Our work expands the
EST model by also investigating the influence of tech- Fig. 1. The systems of Bronfenbrenner’s orig-
nological factors to account for the ways in which inal Ecological Systems Theory (EST) model
modern PI systems’ data and usage can mediate layers [14], which we expand and operationalize to
of the social ecology. Additionally, while prior EST the context of serious mental illness.
research has typically focused on the negative impacts

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:7

of technology on individual development (e.g., violent media content, sensory overload, and atten-
tional problems), we examine potential benefits that personal data can have as well.
Overall, our research is focused on understanding how this model plays out in the context of
PI-supported treatment and management of bipolar disorder, with the goal of better representing
and supporting the diversity of relations and roles in the social ecology surrounding the self-
management of SMI. This empirical work will help us to expand and operationalize EST theory
in ways specifically tailored to the development of future PI systems — collaborative computing
architectures, data representations, and interface designs to better enable self-knowledge, self-
awareness, and collective support in this complicated and important health context. We expand on
the details of our application of EST in the discussion.

4 METHOD
In order to identify and better understand the personal data practices of individuals with bipolar
disorder, their inner circle of care providers, and the wider social ecologies that influence their
self-management experiences, we employed a combination of surveys, interviews, focus groups,
and a visual elicitation activity to engage with participants and members of their social support
networks. Due to the potentially sensitive nature of these practices, we worked closely with our
Institutional Review Boards (IRBs) to ensure that our data collection fell within an ethical and
acceptable scope for all concerned. This included minimizing paperwork through verbal consent
and careful screening of participants to exclude those in mental health crisis or suffering from
comorbid conditions like post-traumatic stress disorder (PTSD) that could be triggered by interview
questions. We also relied solely on self-reported details about health conditions and history.

4.1 Study Procedures and Participants

Our research spanned the summer and fall of 2017 at two different sites, using complementary
protocols: in Study 1, run at the University of Washington, we conducted one-on-one interviews
with individuals diagnosed with BD; in Study 2, run at the University of Colorado Boulder, we
conducted a mix of interviews and focus groups with people with BD and other stakeholders.
Participant demographics from both studies are reported in Table 1 in the paper’s Appendix. Focus
groups offer an opportunity to study individuals in social contexts and in conversation with people
who might have the same or similar experiences, countering assumptions that the individual is the
best or “purest” unit of analysis [102]. Combining one-on-one interviews with focus groups thus
provided a way to contextualize data within a social setting, which supported our goal of learning
more about social dynamics, how individuals with bipolar disorder and their families and friends
cope with the illness, and how data practices might be influenced by social experiences.

4.1.1 Study 1 — Interviews. Participants (P, N =14) were individuals diagnosed with BD, in a
stable state of their illness, and willing to share their personal data practices with us. We recruited
through a solicitation on a website run by a regional health organization, through posts to listservs
for the local chapter of the National Alliance for Mental Illness (NAMI), and using flyers posted
in neighborhood mental health clinics. Psychiatric clinical partners also assisted, including with
question vetting. Screening interviews were conducted over the phone and were used to verify that
participants were over the age of 18, had received a BD diagnosis, had not been hospitalized for
mental health reasons in the last 6 months, and did not have other mental or physical health issues
that would prevent them from sharing their experiences with us (trauma, cognitive impairments,
etc.) Participants’ ages ranged from 20–64 (mean of 46 years old); 9 were female, 5 male. A pre-
interview survey asked about general mental health history and technology attitudes and use.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:8 E. L. Murnane et al.

Face-to-face semi-structured interviews allowed us to establish a rapport with participants,

determine the scope of condition-relevant social circles, and ask deeper questions about attitudes
towards technology and self-tracking, personal experiences living with BD, and concerns about
privacy or other issues related to sharing personal data. We also requested that participants bring
examples of any self-tracking documents (including digital and non-digital formats) used to help
manage their illness that they felt comfortable sharing. Finally, we worked through an exercise in
which we asked participants to use a digital tablet to draw a series of timelines representing their
experiences with BD, on both individual and interpersonal levels. This visual elicitation was used
to clarify, probe, and augment the personal narratives described in the first half of the interview.
Each interview lasted approximately 60 minutes.
4.1.2 Study 2 — Interviews and Focus Groups. Participants included both individuals diagnosed
with BD (Primary Participants – PP, N =5) and people they identified as being close to them and
who provided assistance or support in some way around the management of their mental illness
(Secondary Participants – SP, N =3). Not all recruited PP identified or were accompanied by SP.
All participants were over the age of 18, with an average of 37 years old; 6 were female, 2 male.
To recruit, we distributed flyers and postcards both around campus and in local coffee shops,
restaurants, and other social settings. We further disseminated recruitment materials by asking
participants to share information about the study with their friends and family members (e.g., a PP
asking one or more SP to join them in the study or vice versa) and other social networks (e.g., one
interested participant took a stack of flyers to her support group of families of BD patients). We also
advertised in digital settings by sharing flyers to our Instagram, Facebook, and Twitter networks.
In addition, we recruited through a local BD evaluation and treatment center and coordinated
with DBSA (Depression and Bipolar Support Alliance), local NAMI chapters, and local therapists.
Elicitations took place as face-to-face focus groups and individual interviews, depending on the
configuration and scheduling availability of recruited individuals.
The semi-structured prompts for Study 2 were based closely on those used in Study 1, adapting
phrasing and flow to be more suitable for group sessions and adding a handful of questions to
probe issues related to the stigma surrounding BD in more depth. Interview sessions lasted an
average of 48 minutes, and focus group sessions lasted an average of 100 minutes.

4.2 Analysis
For analysis, interview and focus group data were transcribed and de-identified. The set of questions
that were asked of participants in both studies were analyzed together and will therefore be reported
together. We performed an inductive thematic analysis using open coding techniques, where at
least two researchers viewed each transcript and agreed on emergent themes. As we will describe,
Bronfenbrenner’s social ecological model reflected many of the past and present social relationships
our participants identified, leading us to perform a round of focused coding of our data through
an ecological systems theory lens to better understand interrelations among social networks, PI
practices, and technology.

5 FINDINGS
Returning to the idea of relational recovery [86] overviewed earlier, we first describe how our
participants talked about social relations, the roles that other stakeholders play in their experience
of BD, and the ways PI technologies mediate the connections between these relations and the
individual with BD at the center of care. This organization of our findings foreshadows and parallels
the elements of our descriptive model of this sociotechnical ecology that we later formalize in the
Discussion section.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:9

5.1 Social Relations and Roles

To begin, we characterize social relations relevant to long term self-management of BD — that is,
the individual people, groups, institutions, and more broadly encompassing sociocultural ether
with whom individuals with SMI have direct relationships or operate within, for better or worse.

5.1.1 Healthcare Professionals. All participants described their intimate connections with health
professionals, including general practitioners, nurses, psychiatrists, therapists, and counselors.
Forming these relations could be distressing, particularly right after initial diagnosis (e.g., “The first
doctor I talked to was worthless. Never connected with me, never helped me. And I worked with him for
about six months trying to make it happen.” – P1). Participants noted that when these relationships
were frustrating or uncomfortable, it was often because the care professional was too busy, did
not “click” in terms of personality or working style, or that they were dismissive or minimized the
individual’s BD condition (e.g., “I had one nurse in particular the first time I was in there. We didn’t
get along. She actually said to me it would take a miracle for ‘somebody like me’ to get better.” – P4).
Overall, however, participants described a number of positive attributes and experiences from
their professional care experiences, once they found the right “match”. Making this initial connection
was often viewed as a stroke of luck (e.g., “I took a referral from a friend to see [my therapist]. I
stayed with her for 35 years until she retired a couple years ago. So she was a huge– another lucky
break for me to have someone like that in my life.” – P1). To describe positive relations, participants
used words like “amazing”, “enlightening”, and “supportive” or explained that being connected to a
professional caregiver was “one of the best things that happened to me” (P9). In describing the key
ingredients that made such relations successful, participants called out the importance of access,
trust, and stability (e.g., “I trust him more than I’ll probably trust any provider. He was able to be in
close contact. I have the ability to text him... I have that access. It’s a very comforting fact.” – P2). While
prior work has similarly identified that patient–provider trust is key in the successful treatment of
any number of health conditions [58], it can be particularly crucial for those individuals whose BD
is linked with paranoia, a common symptom.
Similarly, given the dynamic nature of BD, having stability in clinical relationships and the
ability to work with the same person for many years was highly valued by many participants.
Unfortunately, maintaining this stability was often not possible, leading to many participants
working with multiple doctors over the years with very different care styles (e.g., “I’ve had legions of
therapists... I’ve been in therapy for roughly 40 years.” – P8). Changing doctors was understandably
upsetting to participants, who explained that it was often due to changes in their health insurance
coverage, status, or network.
Additional coordination was necessary among care providers when such patient “handoffs” oc-
curred. It was also common for participants to work with teams of care specialists who worked
in concerted management of additional health conditions (e.g., ADHD, pain) that are often co-
morbid with BD [54] or lifestyle problems (e.g., nutrition, fitness) that arose as a result of their
BD. For example, P6’s care team included a psychiatric practitioner, nutritionist, acupuncturist,
and a massage therapist in an effort to cope with his additional psychological issues, poor eating
choices, pain, and tension, respectively, that he attributed to his BD. Clear communication was
also mentioned as crucial, especially considering that BD can be associated with disorganized
thinking or other cognitive issues, especially when experiencing symptoms. Several participants
discussed how they developed a “common language” with their doctor over time. Participants also
appreciated the ability to be honest and open with these professional care relations, especially for
those who normally did not disclose their BD to the outside world or who did not have strong
support networks (e.g., “I just needed somebody to talk to. It was like a paid friend because I didn’t
have any of those sisters or friends. She was getting paid just to listen to me.” – P9).

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:10 E. L. Murnane et al.

5.1.2 Inner Social Circles. Beyond health professionals, family and friends also often played
a central role in the management of participants’ BD, although participants tended to be split in
whether they were closer to their families or their friends.
On average, participants strongly valued and relied on their family members for emotional,
monetary, and other forms of instrumental support (e.g., cooking meals, driving to appointments,
and obtaining medication). For these participants, they identified various members of their families
as a “life raft” essential in helping to navigate the initial BD diagnosis or make it through subsequent
episodes. On the other hand, the remaining participants described a much stronger affinity for their
friend networks (e.g., college roommates, childhood friends, and more recent but strongly bonded
friendships). One participant succinctly articulated this choice:

“I rely on nothing from my family with regard to any kind of illness. If I didn’t have the
same DNA as my parents I would not ever have contact with them again. You don’t get to
pick your family is what I’m getting at. But you get to pick your friends. So I think a lot of
people wind up getting much more support because they can choose their friends. They
can choose people that are going to help them.” (PP1)

Participants explained that having known their families from birth meant that a literal lifetime
of negative memories could build up and never be escaped. In contrast, participants pointed out
that while someone could not go find a new family, it was possible to “start fresh” with friends once
their BD had stabilized (e.g., “I had a bad falling out with my old friends... I just f’ed up too much and
they weren’t understanding and they were also– it was kind of a toxic friend group... So then I actually
started making really good friends once I was stable, and those are still my friends.” – P13).
These negative memories often stemmed from the sense of shame participants perceived they
brought to their family, both due to the consequences of erratic symptomatic behavior and simply
from having the label of mental illness attached to their identity. In several cases, participants noted
that their family members “won’t accept”, “won’t acknowledge”, or are “in denial” regarding their
BD, and in turn, provide no support. Dynamics became particularly complex for participants in
families with a history of mental illness, with support, co-dependence, blame, and dysfunction
becoming entangled and multi-directional.
Participants noted that a family member’s prior awareness and attitude towards mental health
would often determine whether they would react to the individual’s diagnosis and need for help
with compassion and support or with embarrassment and rebuff. Participants also explained that
even if they did not have outright poor relationships with their families that it was often simply
difficult to maintain these bonds over time, for instance when moving away for school or work
imposed a geographical distance and physical separation, and in particular due to the nature of
their condition itself (e.g., “It’s very exhausting to be super engaged with somebody and super close to
them... It’s really hard for people with bipolar to maintain just your average basic relationship. It’s like
you really need somebody and then you really don’t care.” – P10). Similarly, most participants would
also get worn down from a sense of guilt associated with overly worrying their families due to
having a mental illness, which could further drive them to disengage. In some cases, when family
situations became too antagonistic or even abusive, participants severed ties altogether.
The sorts of needs that participants sometimes felt were unmet by their families are precisely
the same qualities they sought in close friends: “acceptance”, “compassion”, “forgiveness”, “tolerance”,
and “understanding”. As one focus group participant put it, “I have someone I can talk to about
it that’s not going to judge me. Because [SP1] has never judged me. Ever.” (PP1). Even though not
related by blood, participants recounted stories of close friends that indicated deep nurturing and
protection (e.g., “I had a couple very close friends that were trying to keep an eye on me and were

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:11

literally chasing me around and making sure I was taking my medication, taking alcohol and drugs
out of my hand... trying to get me to go to sleep and sitting there until I did.” – P10).
Beyond individual friends, participants also explained the importance of group-based support in
managing their BD, whether they be BD-specific support or therapy groups; casual groups of friends
with BD who occasionally met to share mutual support; or online groups connected through blogs,
fora, or general-purpose social media platforms. It was interesting to note that some participants
expressed having a sense of connection with the abstract community of all people with their same
condition, “my people”, even though they had never met nor could articulate specific details about
whom those individuals were. This perceived in-group gave participants an encouraging feeling of
not being alone and in some cases even helped them cope with their initial diagnosis, as it was a
relief to know that what they had been experiencing had a name and that there were other people
in the world to whom they could relate (e.g., “I finally felt like, oh my God, there’s a whole community
of people that experienced the same thing that I do.” – P13).
Overall, similar to professional caregivers, stability was seen as a key ingredient in supportive
family and friends. One participant shared that she intentionally sought a husband who was
extremely predictable and staid — her “human metronome” — explaining that she appreciated the
fact that “he’s boring” because “It’s very reassuring. I don’t need surprises and I don’t like surprise
parties. It’s consistently like knowing night follows day.” (P9). Some participants acknowledged that
change was an inevitable part of long-term SMI management (e.g., “You have friends and relatives
that come in and out of your life that are going to be able to handle it or not, and it takes a while
to get a caregiver support group that meets your needs.” – P6). In some cases, change could be in
a positive direction (e.g., “I actually am okay with the level of support I’m getting from my family
now. They’ve grown an understanding of the disorder.” – P2). However, most sorts of instability were
perceived as quite distressing, for instance due to the death of a family member, a divorce, or a
break up. Also similar was the sense of feeling “fortunate”, “lucky”, and “blessed” to have long-term,
stable relations. For two participants, animals and pets were called out as additional examples of
comforting consistent relationships in their lives that helped keep self-management on track.

5.1.3 Outer Circles and Sociocultural Backdrops. We identified a number of more peripheral
social units that also impacted participants’ BD management, including places of employment
(bosses, coworkers) or education (teachers, classmates, school administration), local community or
religious organizations, and medical insurance groups.
While employment benefitted participants’ self-management in terms of income, access to
healthcare, and overall empowerment, it was also noted as a source of distress due to the difficulty
associated with finding work and the fear of losing existing or prospective employment if their SMI
condition was discovered (e.g., “In my world, you don’t talk about it for purposes of wanting to be
in the working world.” – PP2). Participants also had mixed experiences with local community and
religious organizations, which for some provided a sense of structure and safety; however, becoming
disconnected (e.g., due to a move) from such “sheltering and nurturing communities” (P9) could be
destabilizing and trigger symptoms. Further, the conservative perspectives of military or religious
communities in which a few participants grew up were noted to have approached their early signs
of mental illness with denial, which left participants with a sense of rejection and uncertainty,
rather than a sense of how to constructively move forward with proper treatment. Identified as
another undermining influence, a number of our participants discussed health insurance with us
(namely, losing it or the inability to afford it), indicating how insurers could indirectly enable or
disable condition management. It seems this lack of access to medical care was in many cases the
reason behind why participants reached for technology as part of self-managing their SMI.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:12 E. L. Murnane et al.

Finally, it was clear throughout our conversations that societal attitudes surrounding mental
health had left deep impressions on all of our participants. SMI, and the well-recognized stigma
associated with the condition [22], had largely been absorbed into participants’ identities, who
pointed out that this could make self-compassion and self-care a challenge when they viewed
the self from this demeaning external perspective. One participant, who was of African descent,
specifically pointed out racial differences she had experienced in terms of not only the aversion but
even the acknowledgement of SMI as a legitimate condition (e.g., “My boyfriend just says [about
my aunt’s depression and BD], ‘That sounds like white people problems.”’ – SP3). Such biases also
motivated a general tendency for participants to conceal their SMI. Encouragingly, however, a
number of participants had come to shed these societal attitudes, for instance to embrace a healthy
identity distinct from their illness (e.g., “I’m managing an illness, instead of being bipolar” – PP5) or
to be open and vocal about BD in an effort to increase awareness, humanize the condition, and
ultimately lift stigma and support others in the broader SMI community.

5.2 Personal Data as a Relation and a Relational Mediator

As participants navigated these spheres of social relationships, we identified that the entity con-
stituted by personal data itself could also resemble a kind of social relation for them, in that they
regularly interact with and are influenced by these reflections of themselves, as part of managing
SMI. In addition, this personal data entity could play a mediating role among an individual’s other
social relations identified above.

5.2.1 Personal Informatics Practices and Perspectives. Here we describe how our participants
engage with personal informatics (PI) as part of supporting informal and long-term condition
management. Our findings align with those from prior work [51, 76], which indicates the repre-
sentativeness of our study samples, though we add new insights regarding the ways that many of
these PI practices and attitudes are socially constructed.
Participants made use of a number of different PI methods and mediums, both analog and digital,
including post-it notes, lists, journals, hand-drawn charts and graphs, calendars, spreadsheets,
paper-based or online BD forms (e.g., from the DBSA), smartphone apps, and activity trackers (e.g.,
pedometers, Fitbit). Tracking could also happen invisibly by taking mental notes or by entrusting
tracking to close social connections who had the ability to keep an eye on an individual’s behavior
to watch for changes, notice warning signs, or help fill in memory gaps.
Participants used these methods to track a variety of items — foremost mood — along with
medication adherence and dosing, sleep, exercise, weight, spending, and alcohol/drug/tobacco
use, which were all identified as behavioral proxies of symptom severity or an oncoming episode.
Participants also self-tracked time, noting the value in sticking to routines as a way to minimize
symptoms but also because this made it easier for others to help monitor for erratic behavior (e.g.,
“It’s very much planned. ‘This is where I’m going... if you don’t hear from me by four o’clock, you gotta
call me.”’ – P12). Similarly, social events were often tracked, not only as a way to monitor healthy
social rhythms and avoid social isolation as part of BD management, but also as an opportunity to
notice and celebrate adhering to social commitments, which can be challenging to maintain during
those phases of BD marked by depression or paranoia.
Participants had been engaged in PI for various spans of time, ranging from months through
to fifteen years (P1). In addition, PI records were kept at various levels of tracking granularity.
For participants that tended towards keeping highly detailed data, they noted that while this
was valuable as a concrete representation of their experiences, the associated burdens could be
“overwhelming and confusing” (P2) or “tedious and difficult to manage” (PP4). For some, this eventually
led to abandonment of their PI regimen. Participants also described tradeoffs between digital tools

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:13

being easier to manage but also harder to customize (e.g., by using colored pens or stickers to
annotate paper-based tracking materials with personally meaningful schemas).
Just as participants had varied experiences with particular types of social relations (e.g., positive
vs. negative family relations, etc.) participants were mixed in their overall receptivity toward PI as
well. For instance, some considered themselves natural self-trackers, self-identifying as the tracking
“type” or inherently “inclined” to engage with PI; accordingly, these participants typically found PI
beneficial and often self-initiated tracking outside of clinical care. In contrast, other participants
were strongly turned off by the premise of PI (e.g., “I have never done [self-tracking] because I find it
repugnant... I’m already objectified enough – why would I want to do it to myself?” – P3). Similarly,
some participants valued PI tracking artifacts as an archive and a means to support future self-
reflection about their BD management, whereas others felt such traces solidified evidence of their BD
hardships and failures or led to increased rumination. Another parallel between PI engagement and
social relations is again the importance of stability in the interaction, with participants explaining
how technological problems (e.g., software crashes, lost data, an update removing relied upon
functionality, etc.) could be destabilizing, particularly when it disrupted self-tracking routines that
they had come to strongly depend on for BD management.
5.2.2 Interpersonal Engagement through Personal Informatics. Beyond identifying these ways that
individuals with BD engaged with PI, we also saw how PI could mediate participants’ engagement
with their social relations and those relations’ roles in long-term BD management. In particular, our
findings speak to choices and behaviors around collaboratively collecting, sharing, and interpreting
personal data.
Participants described a few different co-tracking scenarios and motives. Sometimes, a social
relation would take over the capture of personal data (e.g., about medication use or observable
symptoms) when the individual was unwell and therefore unable to self-track, yet the information
still needed to be captured to maintain the continuity and usefulness of the record. In other cases,
collaborative tracking activities (e.g., maintaining a shared notebook or digital calendar) were used
to aid co-reflection and build mutual awareness about how the person was experiencing BD and
to help the social relation “see in” to what the participant might be going through. Despite the
potential benefits of such collaborative tracking activities, the deliberate choice was sometimes
made to avoid co-tracking, in the interest of enhancing the participant’s agency (e.g., “I think [PP4]’s
been great at tracking information. I think it was important that she was the one tracking it and
keeping control of her own ship rather than me or somebody else stepping in. I mean, it just doesn’t
seem right for her.” – SP2).
Further, an interaction among participants with BD in one of our focus group sessions highlighted
how tracking, even if performed on one’s own, can still impact interpersonal awareness — although
in this case, in a problematic way, as the conversation identified how current self-tracking interfaces
can reinforce a sense of normative social comparison:
PP1: Pushing it into categories is so difficult to know: are you high, medium, or low/high?
What does that mean?!
PP2: Is that normal? Is that how other people feel?
PP1: It also makes me feel like everyone else has such a different perception of how the
world works. That can be further alienating in some ways.
While assessing oneself in terms of coarse grained buckets can be difficult for people self-
tracking any variety of personal data, prior research has identified that individuals managing SMI
are particularly vulnerable to feeling judged, flawed, and isolated when confronted with normative
PI measurement scales and visual representations that do not align with their own mental models
that they use to gauge and make sense of their condition [69].

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:14 E. L. Murnane et al.

When it came to sharing personal data, participants varied in their behaviors and preferences.
Sharing with health professionals was typically seen as standard and acceptable, largely in order
to support better patient–doctor communication. Particularly notable is how participants felt PI
sharing could put them on a more equal footing with their clinician, including to push back during
disagreements (e.g., about whether a medication should be adjusted or avoided), because the data
served as a resource that could be called on to support their point of view. For some participants,
simply knowing that an experienced person would be looking at and making sense of the data could
lift a sense of PI burden, given that it could otherwise become overwhelming to feel they were solely
responsible for managing, interpreting, and deciding how to act upon collected information. One
participant mentioned using a self-tracking app that, with permission, would automatically email
data to a clinician. However, when participants experienced uncertainty about whether doctors
actually reflected on shared data, it left a reduced sense of confidence, both about the value of
self-tracking and about the quality of their care relationship.
With family, tracking data was sometimes shared to improve loved ones’ ability to notice if
the participant had forgotten to take medication or overlooked some other aspect of condition
management, in which case they could provide prompts or reminders — or if deemed necessary,
more extreme protective measures could be taken. One participant described an explicit, detailed
“emergency plan” that he had constructed together with his therapist and sisters: “If I start pacing a
lot, if I want some money, if I start drinking again. That’s stage one. Stage two - I’m just not paying
attention to you, still pacing, and I’m giddy - that’s when you need to take me to the hospital.” (P12).
Another participant explained how she had similarly “taught” a new but trusted friend how to
monitor and interpret her online social networking in order to notice behavioral patterns that
forebode an episode. In other cases, data was shared as a way to foster empathy and understanding
from a family member or friend or in an attempt to explain erratic behavior. A similar desire to build
compassion could motivate participants’ decisions to share data with wider social layers to whom
they would not normally have contact in everyday life. For instance, some participants reported
using online social networking platforms as a means of sharing PI outputs and establishing or
strengthening social connections.
Sometimes, however, sharing was avoided precisely to limit the chances of making such connec-
tions — for instance, one participant thought that sharing data could make it easier for her abusive
ex-husband to find her. Relatedly, concerns around privacy and data security (and the consequences
of leaking sensitive health data) were the most common reasons noted by participants for disliking
or altogether avoiding tracking in general (e.g., “I think I ended up being too paranoid to continue
in tracking. Because somebody is going to find my log and it’s going to be really dangerous for me.”
– PP2). In some cases, these concerns were not only self-centric but aimed to protect relations as
well (e.g., “I threw [my tracking journals] away because I was afraid that someday I would die and
people would find them and see things about themselves.” – P10). Destroying data and other tracking
artifacts was a common solution for mitigating perceived risks for oneself and others that might
arise if the data were exposed. For others, such danger seemed low, as they felt it would be difficult
for relations to decipher their personal data without their involvement.
Another observation to note is the way that personal data can be intended only for personal
use but serve socially-facing goals such as impression management. For instance, one participant
described how he would use his tracked information to gain enough self-awareness to identify his
“bounds” (of noticeable versus undetectable symptomatic behavior) in order to then conceal the
presence of his BD and maintain an outward appearance of “normalcy” (e.g., “But I decided pretty
early on if I could maintain in this range right here, I’d have a life, no one would ever know. And that’s
pretty much what I did.” – P1).

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:15

Overall, we saw that participants have diverse perspectives regarding disclosure and complex
criteria when determining which data will be shared, with whom, and how [35, 59]. These orienta-
tions – which fall on a spectrum of openness – can influence a person’s PI attitudes, practices, goals,
and, in turn, the ways that PI can mediate social relations. Where a person falls on that spectrum
can be informed by baseline PI literacies as well as positive or negative life experiences, including
with PI, their BD condition, and their social relations.

6 DISCUSSION
6.1 Personal Informatics, Collective Engagement
At the core of this paper is the idea that tracking activities and data practices previously considered
“personal” are evolving to include — or perhaps more accurately, are revealing themselves to be —
collective phenomena, embedded in multidimensional social structures. The socially embedded
practices we highlight here are related to sensemaking in organizations [101], collective intelligence
[67], distributed cognition [43], and collaborative knowledge production [103]. From this body of
literature came the original notion of “collective sensemaking” as a way to describe the activities
and interactions that comprise these collaborative practices. In this context, personal data can be
seen as a form of boundary object [96], marking the intersections of different social relations in a
PI ecosystem.
Seminal PI literature argues that system design should be approached holistically — taking into
account the various facets of individual lives and supporting iteration between stages of PI use [57].
Other healthcare technology researchers have advocated for a similarly inclusive approach that
considers the overlapping sociotechnical dimensions of personal health management, stresses the
importance of lived experiences and situated practices, and incorporates a variety of stakeholders
[2, 4, 25, 30, 39, 40, 49, 84].
In this vein, our work has identified some of the primary mechanisms by which social rela-
tions in our use case influence decision making, empathy, communication, and mental health
self-management. Our findings indicate that improving the ways that PI systems support collective
engagement can reduce the huge societal costs of long-term treatment of chronic conditions by
(a) providing a range of different types of stakeholders with appropriate access to personal
data records that might facilitate more effective long-term management of chronic conditions,
(b) scaffolding existing social relations at multiple levels within support networks that have
been shown to play important roles in the management of chronic conditions, (c) creating so-
ciotechnical platforms for collective and collaborative practices of gathering, validating,
and interpreting behavioral trace data (including contextually sensitive data models, automated
and semi-automated tracking tools, and visualizations that more closely align with the lived ex-
periences of users), and (d) responding to dynamic changes in social relations and contexts
over time in recognition that longitudinal use is inevitably accompanied by reconfigurations of
trust and agency.

6.2 A Model of the Sociotechnical Ecology Surrounding SMI Management

Our findings inform the creation of an ecological model for personal data practices that encompasses
the range of people, organizational units, and broader sociocultural norms and values that influence
and are influenced by the condition and care of a central individual. Our model is illustrated
in Figure 2 and described in the following subsections. We leverage Ecological Systems Theory
(EST) [14, 15] in creating this model of the social contexts described by participants as playing
an important role in the long term management of their mental health. In doing so, we highlight
key sociotechnical implications and requirements of conceptualizing personal health informatics

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:16 E. L. Murnane et al.

through a collective lens. Our model is descriptive, explanatory, and prescriptive in that it offers
conceptualizations and invites further inquiry, proposes explanations of observed relationships
and processes, and helps inform the design of novel user experiences [90].

6.2.1 System Layers. In the center of our model is the individual managing SMI, along with
the person’s identifying attributes (e.g., age, gender, other demographics, etc.) and condition
characteristics (e.g., type of bipolar disorder, symptom severity, time since diagnosis/undiagnosed,
etc). Surrounding the individual, we define nested “layers” of relations, which reflect their level of
interaction with the person’s condition management.
The microlayer consists of the closest relations that are most directly involved in SMI care.
Based on our dataset, this typically involves the individual’s professional caregivers (clinicians,
nurses, etc.) as well as inner social circles of family, friends, and peer support groups, who together
share the bulk of responsibility in monitoring and facilitating the health of their patient/loved one.
Other work has similarly identified immediate family and very close friends as constituents of
“intimate” care networks that patients involve in health monitoring [39]. For our participants, the
care practices of microlayer relations often involved keeping track of symptoms and medications,
providing emotional support, maintaining a general awareness of an individual’s current stability,
and potentially intervening to adjust behavior or to get help in a crisis. However, we found that
participants’ relations in this layer could also be responsible for exacerbating symptoms or generally
worsening one’s state of health and well-being — for instance, due to dissatisfactory clinical care,
family disharmony (e.g., denial, shame, rejection), or toxic friends who goad triggering behaviors
or are otherwise bad influences.
Next, we define the exolayer as the wider institutional entities that impact how an individual
manages his or her condition. For our participants, these relations included workplaces, schools,
community organizations, and health insurers, which engaged less immediately with them but
could still exert a potent influence by aiding or obstructing instrumental (e.g., financial, medical)
and expressive (e.g., emotional support) needs — for instance, when office politics or admissions
criteria made participants feel pressure to conceal their SMI or when changes to a spouse’s health
insurance affected their own coverage. As with the microlayer, we observed that interactions with
exolayer relations can be beneficial or problematic and that dynamics can and do vary over time.

Fig. 2. The layers and interactions of our ecological model for conceptualizing the social and personal
informatics entities that influence and are influenced by serious mental illness (SMI).

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:17

The macrolayer in our model reflects the individual’s larger sociocultural context — ideologies,
cultures, and societal attitudes — around mental illness. In our case of SMI, this layer also captures
how an individual is impacted by current economic trends as well as government policies regarding
healthcare. While this layer is the broadest and most in-the-background compilation of relations
to the individual, our data illustrated how it can still have a tremendous impact on condition
management as well as identity orientation, typically by minimizing participants’ capacity for
self-compassion and acceptance due to stigmatic societal tropes — although, we did observe that
over time and with age and SMI experience, participants could come to cast off this cultural baggage,
often in large part due to more self-affirming, positive interactions with supportive relations in
their microlayer.
Beyond these nested social layers, our model also contains a cross-cutting temporal layer
that represents patterns and transitions throughout the course of the individual’s life, along with
socio-historical events. Prior theoretical research on collaborative work in healthcare has similarly
identified temporal rhythms and trajectories in medical work (e.g., in shift changes, morning
rounds, medication administration, employee turnover, etc.), although the focus has been on the
professional caregiver perspective and clinical workplaces [23, 88]. These dynamics are particularly
salient in a mental health context, considering the inherently periodic and repeating biological,
psychological, and social fluctuations that underpin an SMI like bipolar disorder [36, 69]. Indeed,
in our dataset we observed a number of common transitional experiences (e.g., changing doctors,
switching medications, death or other traumatic loss, divorce, losing friends, making amends,
discovering support groups, moving house, shifting jobs, and so on) that impacted or were impacted
by participants’ BD. And while many people encounter such flux throughout life, our participants
generally noted that they believed the transitions they experienced were more frequent, extreme,
and/or irrevocable as a result of their condition.
Finally, we include an informatics layer in our model to capture the relational and mediating
roles personal data play in socially-embedded SMI management, as described in Section 5.2. In our
study, we observed that PI artifacts and the representation of self embodied by captured data can
themselves resemble a relational entity, which exists together with the individual at the center
of the ecology. We also saw that analog instruments (notebooks, stickers, paper forms, etc.) were
geared more toward use at the personal or microlayer level, with practices either entirely private
or shared only with close ties that participants lived with or regularly depended on for care. On the
other hand, digital technologies (self-tracking software, online platforms, etc.) enabled interactions
with more indirect layers; however, more risk was associated with data captured this way, especially
if participants had antagonistic relations at the exolayer (discriminatory workplaces, hostile ex-
partners, etc). Further, considering personal data from a macrolayer perspective demonstrates how
PI can act in service of both positive and negative cultural values (e.g., mobile self-tracking apps
affording broader access to care for a vulnerable population or normative measurement scales
reinforcing users’ perceptions of societal stigma).
In addition, the informatics and temporal layers of the model can interact as well, given that
PI attitudes and practices can evolve over time. For our participants, common changes included
tracking more or less based on their level of understanding or progression of their illness, on shifting
perceptions surrounding perceived threats to privacy, or in response to discovering new peer support
communities where sharing PI could aid individual and collective needs, be they informational or
emotional. Overall, including a PI layer in our model promotes the holistic examination of how
not only current PI instantiations but also future advances in PI (e.g., to support novel formats of
information and in more collective, dynamic circumstances) can impact care practices and social
structures.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:18 E. L. Murnane et al.

6.2.2 Dimensions of Relational Interactions. Our model also characterizes the interactions among
these various social ecological layers of an individual’s life. While EST uses the notion of a “mesosys-
tem” [14] to represent interactions between social entities, our data indicates a richer conceptualiza-
tion is necessary in order to reflect the following aspects of relational interplay that our participants
commonly experienced. Specifically, we define the following four dimensions of relational interac-
tions that occur among entities in the ecology and the individual at the center: valence, intensity,
directionality, and dynamism.
We define valence as a reflection of where on the support spectrum a relationship falls, in terms
of the interaction’s positivity/negativity. We believe these five levels — from deeply supportive, to
moderately supportive, to neutral or absent, to irritating, to physically or psychologically abusive —
characterize well the range of valence experienced by our participants, based on our conversations
and the language they used.
A notion of intensity conveys how weakly or strongly influential a relationship is. On one end
of the intensity spectrum, participants reported potent, direct influence (e.g., a doctor visiting a
patient regularly during hospitalization or a friend arranging an “intervention” to address erratic
behavior), spanning to more passive influence perceived in the periphery (e.g., a spouse refilling the
patient’s medication once a month without much notice or an individual’s implicit perceptions of
their church group’s SMI attitudes). As with other aspects of interaction, we saw that the intensity
of influence could oscillate over time — for example, social support would typically intensify in
moments of crisis yet relax when a person was experiencing minimal symptoms.
Directionality indicates who is influencing whom, among an individual and the relations in his
or her social ecology. For example, EST mostly focuses on the relation-to-individual influence (e.g., a
caregiver influencing care). In our study, we frequently observed influence in the opposite direction
as well, with an individual’s condition and self-management routines considerably impacting
others, particularly close relations in the microsystem. Further, while this influence is often uni-
or bi-directional (e.g., between a patient-caregiver dyad), we saw a number of instances of multi-
directional influence as well (e.g., among collaborative care teams, when a child’s diagnosis impacts
the family’s priorities, and when a peer support group comes together).
Finally, the dynamism of engaging with a relation represents whether this relationship or its
influence tends to be highly stable (e.g., a lifelong friend or an entrenched cultural norm) or highly
unstable (e.g., frequent employment turnover for someone whose condition imposes significant
functional challenges). We also saw that participants could experience a “dynamic stability” closer
to the middle of this spectrum; for example, one person stayed for years with the same, specific
clinician — but that clinician regularly fiddled with her medication or treatment plan, which
sometimes led to uncertain outcomes.
Throughout our discussions with participants, we observed that participants encountered diverse
combinations of these dimensions when interacting with relations in their social ecologies, although
some co-occurrences were more common than others. For instance, participants frequently dis-
cussed receiving (directionality) deeply positive support (intensity, valence) from long-term, stable
(dynamism) friends and family relations from their microlevel. An interesting future extension of
this work would be to more thoroughly investigate which layers of the relational ecology tend to
correspond more or less to which dimensions of interactions among those layers. We return to this
point in our discussion.

6.3 Design Implications

The lens of EST provides a means for operationalizing the complex and dynamic social relations
that exist in and around the long-term management of SMI. This is essential when designing a
collective PI system not just within this space, but in many other situations where longitudinal use

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:19

of personal data is entangled in social practices. Here we offer design implications that surfaced as
a result of viewing our use case through EST. Some are specific to SMI contexts, while others are
more general.
Design implication 1: Systems need to accommodate the emergent and dynamic prac-
tices of accruing, breaking, and changing social ties. Changes to social structures — even
those that seem “permanent” like family relations — are inevitable, and the impacts of these changes
can be unpredictable. For example, switching to new clinical caregivers carries the heavy burden
of having to repeatedly recount personal history and assess trust and privacy issues. Access to
both personal health information and insights into how a person is really feeling on a day-to-day
basis are continually re-negotiated by the individuals at the center of these networks of care. This
dynamic management of disclosure [56] needs to be supported as a first-class organizing principle
of these systems. Role-based permissions and user scenarios are not adequate because issues of trust
and authority are socially constructed and context-specific, which means they are also dynamic.
For example, in prior work, researchers developed an interactive visual tool for sharing and jointly
making sense of personal tracker data [87]. In this tool, privacy settings were implemented that
allowed per-recipient access rights to be set so that users would be comfortable with how different
types of PI data would be shared to different members of their social circles. We see this as a
baseline requirement for these kinds of systems — a much different approach than is typically
taken in commercial, “social-sharing” focused PI systems. However, as seen in similar privacy
management contexts like social media, interfaces to facilitate access can quickly become unwieldy
[24, 41, 65]. Further, even similar kinds of relations (e.g., patient–provider relationships) can exhibit
a significant amount of variation and dynamism: one user’s relationship with a clinician might
bear little resemblance to another user’s relationship with a similar type of professional; PI systems
must therefore be cautious in implementing disclosure capabilities using stereotypical role labels
and disclosure or sharing permissions.
Design implication 2: People find value in comparing their data to others’, in many
cases due to the desire to determine baselines and make normative comparisons with
others with whom they feel an affinity. While other apps have been designed to support social
comparison (e.g., of fitness data [93]), such comparisons can be highly problematic in an SMI
context, given the aforementioned risks related to self-scrutiny and unrealistic expectations to
which people with BD are vulnerable. Regardless of the specific representations or visualizations
employed in health-centric PI tools like these, the impact of social, cultural, and normative framing
is particularly important to consider in the context of supporting vulnerable populations, given the
normative expectations inherent in and underlying most existing personal informatics systems [62].
Prior work on social PI systems has found that leaderboards or other interfaces that allow for the
direct comparison of multiple individuals’ data can be motivating due to the sense of competition
induced [47, 51, 73, 92]. However, managing chronic conditions like SMI is not a race to a goal post
with clear-cut measures of success (e.g., more steps are better). Rather, it is a long-term, day-by-day
balancing act of “staying in bounds.”
Careful consideration needs to be given to the social consequences and contracts around sharing.
A failure to design without social interaction in mind brings consequences including limiting
the types of information that can be shared (e.g., less support for exchanging rich experiential
knowledge rather than just statistics), the level of control a user has over that sharing, and the
ability to retroactively review and revise disclosed content [53].
Design implication 3: For people managing SMI, self-tracking is about crisis mitigation
and crisis management. While some individuals talked about the ways in which they enjoyed
or were drawn to personal record-keeping, many also described the vulnerability associated with
recognizing potentially troubling patterns in those traces. Participants described a process of

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:20 E. L. Murnane et al.

spotting a problematic change in a routine or daily behavior and going directly to a healthcare
provider to adjust medication or seek alternative therapies. Others talked about relying on trusted
friends or family members to provide perspective during those times when they could not trust
their own observations or interpretations of personal data. An impending crisis was not typically
signalled by just one event, but by a series of cues. Some participants talked about ways that they
masked or hid potentially dangerous shifts in mood by only exposing friends and loved ones to
certain behaviors. Others talked about the ways that they compared sets of tracked “variables” (our
word, not theirs) in order to make sure that they were getting an accurate picture of how they were
doing (e.g., an unusually large pile of boxes from Amazon on the doorstep plus a series of sleepless
nights indicates that a change for the worse might be imminent). PI tools have the potential to
support triangulation of this nature, both across behavioral and mood markers and among trusted
stakeholders.
Further, participants described their tracking behaviors as being dependent on a number of
factors including their phase of illness, comorbid conditions such as OCD or PTSD, strength and
nature of social support networks, access to formal healthcare in the form of regular therapy and
medications, and ready access to urgent care for mental health. In these cases, adherence to a
tracking regime is not just about the interactivity a tool provides; it can also be a signal to be
tracked in its own right and a significant measure of well-being (e.g., less tracking may correspond
to decreased urgency). This recognition corresponds with a need for situated methods to articulate,
measure, and evaluate use in ways that better reflect holistic PI practices, such as cases where
non-tracking (or syncopated tracking) could be seen as a desirable outcome in conjunction with
other factors.
Design implication 4: PI systems for collective engagement carry the potential of un-
intended consequences, related to both introducing and limiting access to personal data
and particularly within the challenging and dynamic social context of managing SMI. In
our model, some of the mechanisms or trajectories of these consequences can be traced back to
the interactions among levels (e.g., valence, intensity, direction, dynamism). Research and develop-
ment in this space will need to involve careful application of participatory design techniques that
enable the user-centered identification of how to best support dynamic information sharing and
data-access needs required in this context. Our participants shared stories of episodes in which
tracking practices that seemed to offer social benefits in one context corresponded to risks or
harms in another (e.g., being willing to share financial and spending data with a trusted clinician
but guarding that same information from loved ones who might become alarmed, judgmental, or
reactionary).
Just as EST enabled us to operationalize social relations, approaches like Grounded Design [91]
and other participatory methods can help to position PI systems as elements within a dynamic social
context rather than as solutions to a single, static, “solvable” problem. According to Rohde et al. [91],
empirical design research is inherently entangled with social practice. Research activities exist in an
environment that is dynamic, contingent, unpredictable, and organic; this means that the problem
that we set out to solve has often transformed by the time we implement a system. Techniques
like Grounded Design focus on cultivating a shared language for communicating with community
members and a functional understanding of the social practices and contexts to be improved [91].
This requires close collaboration between researchers and community members, which is especially
important when working with vulnerable populations and in light of the complex social dynamics
underlying the management of SMI, as reflected in our model.
After incorporating participants’ visions and values into exemplars of this new class of collective
informatics systems, appropriation studies [91] can be undertaken to examine how they fare when
deployed into the “wild” and incorporated into existing and new social practices by groups of

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:21

real-world stakeholders. This will enable designers to empirically assess the impacts of various
interaction and visualization techniques. Conducting appropriation studies over extended periods
of time will also be necessary in order to capture changes in the course of one’s illness or social
circumstances — and the particular constellation of social relations in which that individual’s
condition is being managed.
Additionally, if researchers and practitioners are going to take on the challenge of delineating PI
ecosystems in terms of multi-tiered social relations — one of the primary outcomes that we make
in this work — then we also have to acknowledge that these tools and practices are unfolding in a
sociotechnical context where concerns about privacy, identity theft, surveillance, and manipulation
are rampant. Many who work in the area of PI are well aware of the sensitivity of personal data;
however, significant work in data science is being done to draw ever more refined representations
of people and their behaviors from larger and more complex data sets [1]. This work creates a
moving target for ethical standards, raising issues of algorithmic inclusivity and normative data
models, as well as questions about what consent looks like in our current data landscape. Building
on the work of Bietz and others [12, 18], our findings reinforce the need for PI systems designers to
be reflective about how tools like automated sensors can catapult a self-tracking app into a data
security design space and require more sophisticated models and awareness of the data literacies
of end users.

6.4 Limitations and Future Work

In this study, we focused primarily on the patient perspective and the rich attitudes and experiences
these individuals hold surrounding their social recovery relations, their use of PI, and their chal-
lenges and needs when it comes to engaging others with and through their personal data. While we
were able to include a few family and friends in our focus group study, it will be important going
forward to interact with additional relations, including members of the exolayer. Such future steps
will not only help validate our claims and refine our proposed model, but they will also ensure that
additional relations’ perspectives are surfaced and considered during the PI design process. Further,
there is value in engaging with relations whose interactions exhibit additional dimensions that
we identified in our model; for instance, while we worked with secondary participants who were
relatively stable relations of positive valence and mostly unidirectional influence, it is also neces-
sary to have discussions with relations of more negative valence or that have a more codependent
relationship with the primary participant in order to better understand their point of view and how
novel PI systems might work to repair or support those interactions. Further, identifying potential
points of tension between the preferences of patients and those relations who have the ability to
strongly and positively influence the patient’s SMI management will also be important to balance
the needs of these diverse user groups.

7 CONCLUSION
Focusing on the case of serious mental illness and bipolar disorder specifically, this paper explored
the multidimensional sociotechnical landscape that individuals navigate as part of the long-term
management of their condition. In doing so, we provided empirical and theoretical support for how
personal informatics systems can better align with the interpersonal needs and values of users,
their lived social experiences, and their contexts of relational recovery. In particular, these insights
allowed us to formulate an ecological model that conceptualizes the social relations that play a role
in self-management, along with the valence, directionality, and other dynamics of those interactions.
Finally, we contributed broadly applicable design guidelines that help define an emerging class of
collective informatics infrastructures and interfaces aimed at supporting the social ecologies of
personal data practices.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:22 E. L. Murnane et al.

ACKNOWLEDGMENTS
This research was supported by the University of Washington Royalty Research Fund (RRF) under
Award Number A118704. Additionally, we thank our participants for generously sharing their
experiences and entrusting us with their stories. We would also like to acknowledge the foundational
work of Mark Matthews in early stages of this research as well as the Consortium for Sociotechnical
Systems Research (CSST) for its ongoing support and encouragement of this work.

A APPENDIX

Study 1 (P) Study 2 (PP) Study 2 (SP)

N 14 5 3
Average age 46 (ranging 20–64) 38 30
Female = 9 Female = 5 Female = 1
Sex
Male = 5 Male = 2
Married w/o kids = 1
Single never married = 4 Single never married = 3 Single never married = 1
Divorced = 3
Marital Unreported = 1
status Separated = 1 Separated = 1
Married with kids = 1 Married with kids = 1 Married with kids = 1
Living w/ partner = 1
In relationship, not living together = 2 In relationship, not living together = 1
High school or equiv = 2
Some college = 4 Some college = 1 Some college = 2
Education 4-year degree = 6 4-year degree = 1 4-year degree = 1
Graduate degree = 1 Graduate degree = 2
Unreported = 1 Advanced degree = 1
Disability = 4
Disabled, works PT = 1
Student = 1 Student = 1 Student = 1
Unemployed = 1 Unemployed = 1
Employment Employed student = 2 Employed student = 2
Retired = 2
Employed part time= 1
Employed FT = 1 Employed, FT = 1 Employed, FT = 2
Unreported = 1
Type 2 = 9 Type 2 = 4
Type 1 = 3 Type 1 = 1
Type of bipolar N/A
NOS = 1 NOS = 1
Unknown = 1
Age of Average = 18 Average = 9
N/A
first episode Range = 12–27 Range = 11–17
How long Average = 10.9 years Average = 9.4 years
N/A
since diagnosis Range = 1–40 years Range = 3–14 years
How long under Average = 15.3 Average = 8.6
N/A
clinical care Range = 0–40 years Range = 3–20 years
Table 1. Summary of participants’ self-reported characteristics.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:23

REFERENCES
[1] Tim Althoff, Rok Sosič, Jennifer L. Hicks, Abby C. King, Scott L. Delp, and Jure Leskovec. 2017. Large-scale physical
activity data reveal worldwide activity inequality. Nature 547 (July 2017), 336.
[2] Eirik Årsand and George Demiris. 2008. User-centered methods for designing patient-centric self-help tools. Infor-
matics for Health and Social Care 33, 3 (2008), 158–169.
[3] Ross J Baldessarini and Leonardo Tondo. 2003. Suicide risk and treatments for patients with bipolar disorder. JAMA
290, 11 (2003), 1517–1519.
[4] Stinne Aaløkke Ballegaard, Thomas Riisgaard Hansen, and Morten Kyng. 2008. Healthcare in everyday life: designing
healthcare services for daily life. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
ACM, 1807–1816.
[5] Jakob E Bardram, Mads Frost, Károly Szántó, Maria Faurholt-Jepsen, Maj Vinberg, and Lars Vedel Kessing. 2013.
Designing mobile health technology for bipolar disorder: a field trial of the monarca system. In Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems. ACM, 2627–2636.
[6] Jakob E Bardram and Mads M Frost. 2017. Double-loop health technology: enabling socio-technical design of personal
health technology in clinical practice. In Designing Healthcare That Works. Elsevier, 167–186.
[7] Irán Barrera. 2008. An ecological systems theory approach in looking at mental health care barriers in the Latino
community. Ph.D. Dissertation. The University of Texas at Arlington.
[8] David Barton and Carmen KM Lee. 2012. Redefining vernacular literacies in the age of Web 2.0. Applied linguistics 33,
3 (2012), 282–298.
[9] Marc Berg. 1999. Accumulating and coordinating: occasions for information technologies in medical work. Computer
Supported Cooperative Work (CSCW) 8, 4 (1999), 373–401.
[10] Andrew BL Berry, Catherine Lim, Andrea L Hartzler, Tad Hirsch, Edward H Wagner, Evette Ludman, and James D
Ralston. 2017. How values shape collaboration between patients with multiple chronic conditions and spousal
caregivers. In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. ACM, 5257–5270.
[11] Joyce Bierbooms, Hans Van Oers, Jeroen Rijkers, and Inge Bongers. 2016. Development of a comprehensive model
for stakeholder management in mental healthcare. Journal of Health Organization and Management 30, 4 (2016),
630–647.
[12] Matthew J Bietz, Cinnamon S Bloss, Scout Calvert, Job G Godino, Judith Gregory, Michael P Claffey, Jerry Sheehan,
and Kevin Patrick. 2015. Opportunities and challenges in the use of personal health data for health research. Journal
of the American Medical Informatics Association 23, e1 (2015), e42–e48.
[13] Dale E Brashers, Daena J Goldsmith, and Elaine Hsieh. 2002. Information seeking and avoiding in health contexts.
Human Communication Research 28, 2 (2002), 258–271.
[14] Urie Bronfenbrenner. 1979. The ecology of human development: Experiments by nature and design. Technical Report.
[15] Urie Bronfenbrenner. 1992. Ecological systems theory. Jessica Kingsley Publishers.
[16] Finn Brunton and Helen Nissenbaum. 2011. Vernacular resistance to data collection and analysis: A political theory
of obfuscation. First Monday 16, 5 (2011).
[17] Yunan Chen, Karen Cheng, Charlotte Tang, Katie A Siek, and Jakob E Bardram. 2013. Is my doctor listening to
me? Impact of health IT systems on patient-provider interaction. In CHI’13 Extended Abstracts on Human Factors in
Computing Systems. ACM, 2419–2426.
[18] Cynthia Cheung, Matthew J Bietz, Kevin Patrick, and Cinnamon S Bloss. 2016. Privacy attitudes among early adopters
of emerging health technologies. PLOS One 11, 11 (2016), e0166389.
[19] Chia-Fang Chung, Elena Agapie, Jessica Schroeder, Sonali Mishra, James Fogarty, and Sean A Munson. 2017. When
personal tracking becomes social: Examining the use of Instagram for healthy eating. In Proceedings of the 2017 CHI
Conference on Human Factors in Computing Systems. ACM, 1674–1687.
[20] Chia-Fang Chung, Kristin Dew, Allison Cole, Jasmine Zia, James Fogarty, Julie A Kientz, and Sean A Munson. 2016.
Boundary negotiating artifacts in personal informatics: Patient-provider collaboration with patient-generated data. In
Proceedings of the 19th ACM Conference on Computer-Supported Cooperative Work & Social Computing. ACM, 770–786.
[21] Ellen P Cook. 2015. Understanding people in context: The ecological perspective in counseling. John Wiley & Sons.
[22] Patrick W Corrigan and Deepa Rao. 2012. On the self-stigma of mental illness: Stages, disclosure, and strategies for
change. The Canadian Journal of Psychiatry 57, 8 (2012), 464–469.
[23] Gavin Doherty, Joseph McKnight, and Saturnino Luz. 2010. Fieldwork for requirements: Frameworks for mobile
healthcare applications. International Journal of Human-Computer Studies 68, 10 (2010), 760–776.
[24] Serge Egelman, Andrew Oates, and Shriram Krishnamurthi. 2011. Oops, I did it again: mitigating repeated access
control errors on facebook. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM,
2295–2304.
[25] Stefan Rennick Egglestone, Lesley Axelrod, Thomas Nind, Ruth Turk, Anna Wilkinson, Jane Burridge, Geraldine
Fitzpatrick, Sue Mawson, Zoe Robertson, Ann Marie Hughes, et al. 2009. A design framework for a home-based

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:24 E. L. Murnane et al.

stroke rehabilitation system: Identifying the key components. In Pervasive Computing Technologies for Healthcare.
IEEE, 1–8.
[26] Daniel Epstein, Felicia Cordeiro, Elizabeth Bales, James Fogarty, and Sean Munson. 2014. Taming data complexity
in lifelogs: exploring visual cuts of personal informatics data. In Proceedings of the 2014 Conference on Designing
Interactive Systems. ACM, 667–676.
[27] Daniel A Epstein, Alan Borning, and James Fogarty. 2013. Fine-grained sharing of sensed physical activity: a value
sensitive approach. In Proceedings of the 2013 ACM International Conference on Pervasive and Ubiquitous Computing.
489–498.
[28] Daniel A Epstein, Bradley H Jacobson, Elizabeth Bales, David W McDonald, and Sean A Munson. 2015. From nobody
cares to way to go!: A Design Framework for Social Sharing in Personal Informatics. In Proceedings of the 18th ACM
Conference on Computer Supported Cooperative Work & Social Computing. ACM, 1622–1636.
[29] Daniel A Epstein, An Ping, James Fogarty, and Sean A Munson. 2015. A lived informatics model of personal informatics.
In Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing. ACM, 731–742.
[30] Geraldine Fitzpatrick. 2011. New challenges for health IT–Design fit for life. In Proceedings of the First European
Conference on Design, Vol. 4. 121–135.
[31] Geraldine Fitzpatrick and Gunnar Ellingsen. 2013. A review of 25 years of CSCW research in healthcare: contributions,
challenges and future agendas. Computer Supported Cooperative Work (CSCW) 22, 4-6 (2013), 609–665.
[32] Ellen Frank, Holly A Swartz, and Elaine Boland. 2007. Interpersonal and social rhythm therapy: an intervention
addressing rhythm dysregulation in bipolar disorder. Dialogues in clinical neuroscience 9, 3 (2007), 325.
[33] Jon Froehlich, Mike Y Chen, Sunny Consolvo, Beverly Harrison, and James A Landay. 2007. MyExperience: a system
for in situ tracing and capturing of user feedback on mobile phones. In Proceedings of the 5th International Conference
on Mobile Systems, Applications and Services. ACM, 57–70.
[34] Mads Frost, Afsaneh Doryab, Maria Faurholt-Jepsen, Lars Vedel Kessing, and Jakob E Bardram. 2013. Supporting
disease insight through data analysis: refinements of the monarca self-assessment system. In Proceedings of the 2013
ACM international joint conference on Pervasive and ubiquitous computing. ACM, 133–142.
[35] Erving Goffman. 1963. Stigma: Notes on a spoiled identity. Jenkins, JH & Carpenter (1963).
[36] Frederick K Goodwin and Kay Redfield Jamison. 2007. Manic-depressive illness: bipolar disorders and recurrent
depression. Vol. 1. Oxford University Press.
[37] Franz Gravenhorst, Amir Muaremi, Jakob Bardram, Agnes Grünerbl, Oscar Mayora, Gabriel Wurzer, Mads Frost, Venet
Osmani, Bert Arnrich, Paul Lukowicz, et al. 2015. Mobile phones as medical devices in mental disorder treatment: an
overview. Personal and Ubiquitous Computing 19, 2 (2015), 335–353.
[38] Kathryn Greene, Valerian J Derlega, and Alicia Mathews. 2006. Self-disclosure in personal relationships. The
Cambridge handbook of personal relationships (2006), 409–427.
[39] Erik Grönvall and Nervo Verdezoto. 2013. Beyond self-monitoring: understanding non-functional aspects of home-
based healthcare technology. In Proceedings of the 2013 ACM international joint conference on Pervasive and ubiquitous
computing. 587–596.
[40] Vigdis Heimly. 2010. How can the Locales Framework be used as basis for design of collaborative systems in shared
health care?. In Collaborative Technologies and Systems (CTS). IEEE, 207–215.
[41] Bernie Hogan. 2010. The presentation of self in the age of social media: Distinguishing performances and exhibitions
online. Bulletin of Science, Technology & Society 30, 6 (2010), 377–386.
[42] Steven Houben, Mads Frost, and Jakob E Bardram. 2015. Collaborative affordances of hybrid patient record technologies
in medical work. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social
Computing. 785–797.
[43] Edwin Hutchins. 1995. Cognition in the Wild. MIT press.
[44] Maree L Inder, Marie T Crowe, Peter R Joyce, Stephanie Moor, Janet D Carter, and Sue E Luty. 2010. “I really don’t
know whether it is still there”: ambivalent acceptance of a diagnosis of bipolar disorder. Psychiatric Quarterly 81, 2
(2010), 157–165.
[45] Maree L Inder, Marie T Crowe, Stephanie Moor, Suzanne E Luty, Janet D Carter, and Peter R Joyce. 2008. “I actually
don’t know who I am”: The impact of bipolar disorder on the development of self. Psychiatry: Interpersonal and
Biological Processes 71, 2 (2008), 123–133.
[46] Maia L Jacobs, James Clawson, and Elizabeth D Mynatt. 2015. Comparing health information sharing preferences of
cancer patients, doctors, and navigators. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative
Work & Social Computing. ACM, 808–818.
[47] Yuan Jia, Likun Liu, Xing Yu, and Stephen Voida. 2017. Designing leaderboards for gamification: Considering perceived
differences based on user ranking, application domain, and personality traits. In Proceedings of the SIGCHI Conference
on Human Factors in Computing Systems (CHI 2017). ACM Press, New York, 1949–1960.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:25

[48] David A Kahn, Gary S Sachs, David J Printz, Daniel Carpenter, John P Docherty, and Ruth Ross. 2000. Medication
treatment of bipolar disorder 2000: a summary of the expert consensus guidelines. Journal of Psychiatric Practice 6, 4
(2000), 197–211.
[49] Elizabeth Kaziunas and Mark S Ackerman. 2015. Designing for lived health: A practice-based approach for person-
centered health information technologies. In Designing socially embedded technologies in the real-world. Springer,
357–381.
[50] Elizabeth Kaziunas, Mark S Ackerman, Silvia Lindtner, and Joyce M Lee. 2017. Caring through Data: Attending to the
Social and Emotional Experiences of Health Datafication.. In CSCW. 2260–2272.
[51] Christina Kelley, Bongshin Lee, and Lauren Wilcox. 2017. Self-tracking for Mental Wellness: Understanding Expert
Perspectives and Student Experiences. In Proceedings of the 2017 CHI Conference on Human Factors in Computing
Systems. ACM, 629–641.
[52] Helen Kennedy, Rosemary Lucy Hill, Giorgia Aiello, and William Allen. 2016. The work that visualisation conventions
do. Information, Communication & Society 19, 6 (2016), 715–735.
[53] Elisabeth Kersten-van Dijk and Wijnand A IJsselsteijn. 2016. Design Beyond the Numbers: Sharing, Comparing,
Storytelling and the Need for a Quantified Us. IxD&A 29 (2016), 121–135.
[54] KR Krishnan. 2005. Psychiatric and medical comorbidities of bipolar disorder. Psychosomatic Medicine 67, 1 (2005),
1–8.
[55] David J Kupfer, Ellen Frank, Victoria J Grochocinski, Patricia A Cluss, Patricia R Houck, and Debra A Stapf. 2002.
Demographic and clinical characteristics of individuals in a bipolar disorder case registry. Journal of Clinical Psychiatry
63, 2 (2002), 120–125.
[56] Edgar Levenson. 1996. Aspects of self-revelation and self-disclosure. Contemporary Psychoanalysis 32, 2 (1996),
237–248.
[57] Ian Li, Anind Dey, and Jodi Forlizzi. 2010. A stage-based model of personal informatics systems. In Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems. ACM, 557–566.
[58] Catherine Lim, Andrew BL Berry, Tad Hirsch, Andrea L Hartzler, Edward H Wagner, Evette Ludman, and James D
Ralston. 2016. It Just Seems Outside My Health: How Patients with Chronic Conditions Perceive Communication
Boundaries with Providers. In Proceedings of the 2016 ACM Conference on Designing Interactive Systems. 1172–1184.
[59] Bruce G Link, Lawrence H Yang, Jo C Phelan, and Pamela Y Collins. 2004. Measuring mental illness stigma.
Schizophrenia bulletin 30, 3 (2004), 511.
[60] Jennifer D Lish, Susan Dime-Meenan, Peter C Whybrow, R Arlen Price, and Robert MA Hirschfeld. 1994. The National
Depressive and Manic-depressive Association (DMDA) survey of bipolar members. Journal of Affective Disorders 31, 4
(1994), 281–294.
[61] Edith Talina Luhanga, Akpa Akpro Elder Hippocrate, Hirohiko Suwa, Yutaka Arakawa, and Keiichi Yasumoto. 2018.
Identifying and Evaluating User Requirements for Smartphone Group Fitness Applications. IEEE Access 6 (2018),
3256–3269.
[62] Deborah Lupton. 2013. Quantifying the body: monitoring and measuring health in the age of mHealth technologies.
Critical Public Health 23, 4 (2013), 393–403.
[63] Deborah Lupton. 2014. Self-tracking cultures: towards a sociology of personal informatics. In Proceedings of the 26th
Australian Computer-Human Interaction Conference on Designing Futures: the Future of Design. ACM, 77–86.
[64] Deborah Lupton. 2015. Lively data, social fitness and biovalue: The intersections of health self-tracking and social
media. (2015).
[65] Michelle Madejski, Maritza Johnson, and Steven M Bellovin. 2012. A study of privacy settings errors in an online
social network. In IEEE International Conference on Pervasive Computing and Communications. 340–345.
[66] Julie Maitland and Matthew Chalmers. 2011. Designing for peer involvement in weight management. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 315–324.
[67] Thomas W Malone, Robert Laubacher, and Chrysanthos Dellarocas. 2009. Harnessing crowds: Mapping the genome
of collective intelligence. (2009).
[68] Mark Matthews, Saeed Abdullah, Elizabeth Murnane, Stephen Voida, Tanzeem Choudhury, Geri Gay, and Ellen Frank.
2016. Development and evaluation of a smartphone-based measure of social rhythms for bipolar disorder. Assessment
23, 4 (2016), 472–483.
[69] Mark Matthews, Elizabeth Murnane, and Jaime Snyder. 2017. Quantifying the Changeable Self: The role of self-tracking
in coming to terms with and managing bipolar disorder. Human–Computer Interaction 32, 5-6 (2017), 413–446.
[70] Mark Matthews, Elizabeth Murnane, Jaime Snyder, Shion Guha, Pamara Chang, Gavin Doherty, and Geri Gay. 2017.
The double-edged sword: A mixed methods study of the interplay between bipolar disorder and technology use.
Computers in Human Behavior 75 (2017), 288–300.
[71] Mark Matthews, Stephen Voida, Saeed Abdullah, Gavin Doherty, Tanzeem Choudhury, Sangha Im, and Geri Gay.
2015. In situ design for mental illness: Considering the pathology of bipolar disorder in mhealth design. In Proceedings

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
127:26 E. L. Murnane et al.

of the 17th International Conference on Human-Computer Interaction with Mobile Devices and Services. ACM, 86–97.
[72] David J Miklowitz, Michael W Otto, Ellen Frank, Noreen A Reilly-Harrington, Stephen R Wisniewski, Jane N Kogan,
Andrew A Nierenberg, Joseph R Calabrese, Lauren B Marangell, Laszlo Gyulai, et al. 2007. Psychosocial treatments
for bipolar depression: a 1-year randomized trial from the Systematic Treatment Enhancement Program. Archives of
General Psychiatry 64, 4 (2007), 419–426.
[73] Andrew D Miller and Elizabeth D Mynatt. 2014. StepStream: a school-based pervasive social fitness system for
everyday adolescent health. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM,
2823–2832.
[74] Sonali R Mishra, Andrew D Miller, Shafli Haldar, Maher Khelifi, Jordan Eschler, Rashmi G Elera, Ari H Pollack, and
Wanda Pratt. 2018. Supporting Collaborative Health Tracking in the Hospital: Patients’ Perspectives. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems. ACM.
[75] Sean A Munson, Hasan Cavusoglu, Larry Frisch, and Sidney Fels. 2013. Sociotechnical challenges and progress in
using social media for health. Journal of Medical Internet Research 15, 10 (2013).
[76] Elizabeth L Murnane, Dan Cosley, Pamara Chang, Shion Guha, Ellen Frank, Geri Gay, and Mark Matthews. 2016.
Self-monitoring practices, attitudes, and needs of individuals with bipolar disorder: implications for the design of
technologies to manage mental health. Journal of the American Medical Informatics Association 23, 3 (2016), 477–484.
[77] Elizabeth L Murnane, David Huffaker, and Gueorgi Kossinets. 2015. Mobile health apps: adoption, adherence, and
abandonment. In Adjunct Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous
Computing. ACM, 261–264.
[78] Christopher JL Murray and Alan D Lopez. 1996. Evidence-based health policy–lessons from the Global Burden of
Disease Study. Science 274, 5288 (1996), 740–743.
[79] Dawn Nafus. 2016. Quantified: Biosensing technologies in everyday life. MIT Press.
[80] National Alliance on Mental Illness (NAMI). 2016. Mental Health By the Numbers. Washington, DC (2016).
[81] National Research Council. 2006. Improving the quality of health care for mental and substance-use conditions.
Quality Chasm Series (2006).
[82] Francisco Nunes and Geraldine Fitzpatrick. 2015. Self-care technologies and collaboration. International Journal of
Human-Computer Interaction 31, 12 (2015), 869–881.
[83] Kathleen O’Leary, Stephen M Schueller, Jacob O Wobbrock, and Wanda Pratt. 2018. “Suddenly, we got to become
therapists for each other”: Designing Peer Support Chats for Mental Health. In Proceedings of the 2018 CHI Conference
on Human Factors in Computing Systems. ACM, 331.
[84] Mustafa Ozkaynak, Rupa Valdez, Richard J Holden, and Jason Weiss. 2018. Infinicare framework for integrated
understanding of health-related activities in clinical and daily-living contexts. Health Systems 7, 1 (2018), 66–78.
[85] Laura R Pina, Sang-Wha Sien, Teresa Ward, Jason C Yip, Sean A Munson, James Fogarty, and Julie A Kientz. 2017.
From Personal Informatics to Family Informatics: Understanding Family Practices around Health Monitoring.. In
CSCW. 2300–2315.
[86] Rhys Price-Robertson, Angela Obradovic, and Brad Morgan. 2017. Relational recovery: beyond individualism in the
recovery approach. Advances in Mental Health 15, 2 (2017), 108–120.
[87] Aare Puussaar, Adrian K Clear, and Peter Wright. 2017. Enhancing Personal Informatics Through Social Sensemaking.
In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. ACM, 6936–6942.
[88] Madhu C Reddy, Paul Dourish, and Wanda Pratt. 2006. Temporality in medical work: Time also matters. Computer
Supported Cooperative Work (CSCW) 15, 1 (2006), 29–53.
[89] Brandon Ellis Rodgers. 2009. An ecological approach to understanding the stigma associated with receiving mental
health services: the role of social proximity. Ph.D. Dissertation. Virginia Tech.
[90] Yvonne Rogers. 2012. HCI theory: classical, modern, and contemporary. Synthesis Lectures on Human-Centered
Informatics 5, 2 (2012), 1–129.
[91] Markus Rohde, Peter Brödner, Gunnar Stevens, Matthias Betz, and Volker Wulf. 2017. Grounded Design–a praxeolog-
ical IS research perspective. Journal of Information Technology 32, 2 (2017), 163–179.
[92] John Rooksby, Mattias Rost, Alistair Morrison, and Matthew Chalmers Chalmers. 2014. Personal tracking as lived
informatics. In Proceedings of the 32nd annual ACM conference on Human factors in computing systems. ACM, 1163–
1172.
[93] Mattias Rost, John Rooksby, and Claire McCallum. 2015. FITtogether: An ‘Average’ Activity Tracker. Beyond Personal
Informatics CHI Workshop (2015).
[94] Jessica Schroeder, Jane Hoffswell, Chia-Fang Chung, James Fogarty, Sean Munson, and Jasmine Zia. 2017. Supporting
patient-provider collaboration to identify individual triggers using food and symptom journals. In Proceedings of the
Conference on Computer-Supported Cooperative Work, Vol. 2017. NIH Public Access, 1726.
[95] Petr Slovák, Joris Janssen, and Geraldine Fitzpatrick. 2012. Understanding heart rate sharing: towards unpacking
physiosocial space. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 859–868.

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.
Personal Informatics in Interpersonal Contexts 127:27

[96] Susan Leigh Star and James R Griesemer. 1989. Institutional Ecology, ‘Translations’ and Boundary Objects: Amateurs
and Professionals in Berkeley’s Museum of Vertebrate Zoology, 1907-39. Social studies of science 19, 3 (1989), 387–420.
[97] Jarrett R Takayama. 2010. Ecological systems theory of Asian American mental health service seeking. (2010).
[98] Paul C Tang, Joan S Ash, David W Bates, J Marc Overhage, and Daniel Z Sands. 2006. Personal health records:
definitions, benefits, and strategies for overcoming barriers to adoption. Journal of the American Medical Informatics
Association 13, 2 (2006), 121–126.
[99] Stephen Voida, Mark Matthews, Saeed Abdullah, Mengxi Chrissie Xi, Matthew Green, Won Jun Jang, Donald Hu,
John Weinrich, Prashama Patil, Mashfiqui Rabbi, et al. 2013. MoodRhythm: tracking and supporting daily rhythms. In
Proceedings of the 2013 ACM conference on Pervasive and ubiquitous computing adjunct publication. ACM, 67–70.
[100] Ron Wakkary and Leah Maestri. 2007. The resourcefulness of everyday design. In Proceedings of the 6th ACM SIGCHI
Conference on Creativity & Cognition. ACM, 163–172.
[101] Karl E Weick. 1995. Sensemaking in organizations. Vol. 3. Sage.
[102] Sue Wilkinson. 1998. Focus groups in feminist research: Power, interaction, and the co-construction of meaning.
Women’s Studies International Forum 21, 1 (Jan. 1998), 111–125.
[103] Stefan Wuchty, Benjamin F Jones, and Brian Uzzi. 2007. The increasing dominance of teams in production of
knowledge. Science 316, 5827 (2007), 1036–1039.
[104] Haining Zhu, Joanna Colgan, Madhu Reddy, and Eun Kyoung Choe. 2016. Sharing patient-generated data in clinical
practices: an interview study. In AMIA Annual Symposium Proceedings, Vol. 2016. 1303.

Received April 2018; revised July 2018; accepted September 2018

Proceedings of the ACM on Human-Computer Interaction, Vol. 2, No. CSCW, Article 127. Publication date: November 2018.

View publication stats