Designing for Dynamics of Agency in NYC Homeless Shelters

Project Context

Navigating Home aimed to understand why the task of assisting NYC shelter residents in finding permanent housing is executed with varying degrees of success across the general population homeless shelter network. The project’s focus was two-fold: understanding what makes some shelters ‘positive deviants’ (Pascale, Sternin, Sternin 2010) that consistently meet their move out targets; and understanding how housing specialists – a shelter staff member who assists residents in navigating NYC’s housing market, understanding subsidies, and applying for housing – might be best leveraged to impact move-out rates.

PPL was enlisted to design and test interventions – such as tools, materials, and processes – that would support both frontline staff and residents in the effort of gaining and maintaining permanent housing. As a non-profit public sector innovation (PSI) lab that partners with government agencies to address challenges faced by low-income and at-risk Americans, PPL combines systems thinking and human-centered research, design, and evaluation methods. PSI is a growing field that aims to innovate public sector service provision through partnership-focused and user-centered practices. The PPL approach to policy and service delivery redesign reflects emerging practices in PSI. As a design firm that positions itself as an operational and strategic partner to government, PPL’s practices are both strategic and material, focusing on system interventions that seek multi-level impact. These include policy changes through to behavior change compelled by new or improved interactions between people, as supported by digital and physical tools. From this perspective, Navigating Home was focused towards producing a ‘proof of concept’: identifying and validating areas of intervention that could improve move-out rates. In order to test the validity of these interventions, PPL collaborated with government partners, shelter staff, and residents to design and test a set of service concepts and related prototypes. The five-month project (December 2018 – April 2019) comprised 250 hours of field research and co-design with 215 shelter staff and residents across twelve shelters in NYC. This Phase One proof-of-concept work laid the foundation for Phase Two, in which program models and tools are being further developed and tested, leading to small-scale pilots.

A Cautionary Note

A project such as this requires thoughtfulness and caution. With so many interdependent variables at play, practitioners must be mindful of the systems they’re designing within and for. Broadly, these risks relate to power and potential unintended consequences of perpetuating unjust dynamics that already limit shelter residents’ ability to seek and maintain housing, as well as frontline staff’s ability to serve them. It is worth noting that this concern extends well beyond Navigating Home, as every social service plays a hand in perpetuating dynamics of power and agency. PSI practitioners need to be highly sensitive to these dynamics, asking: where is power in excess? Where is agency lacking but assumed, meaning that expectations can never be met as people aren’t being empowered to do what’s required or in their best interest? In addition, when considering the circulation of power and agency from a systems perspective, these questions need to be asked not just about the people being served, but also of the government agency client and the project team themselves. As one of PPL’s directors noted, “one of the things we have to do in our work, as mediators between large government systems and individual humans, is to manage power dynamics. We have to mitigate our own power – our own increased power over members of the community, and then we need to turn around and carry those stories into rooms of people who have more power than we do – in and over public systems” (2019).

If these power dynamics are not explicitly addressed and mitigated as a part of design practice, very significant risks relating to the unintended consequences may emerge, given that the service users are often at-risk and in vulnerable, disempowering circumstances. As Paola Pierri argues, “The question of agency is one that cannot be given for granted or ignored, especially when design practitioners are involved in societal issues where dynamics of exclusion and self-exclusion are at play, which can prevent people to act in their own interest…” (2017, S2956–S2957). Therefore, when designing for the social sector, designers must be cognizant of the complex systems that can either impede or aid an individual’s expression of agency.

This article therefore surveys what PPL learned during Navigating Home in terms of working with agency as PSI research and design practitioners and defining a methodology which was sensitive to its dynamics. Rather than presenting a blow-by-blow of each project phase and its outcomes, this article focuses on four areas that were most critical for team reflection and practice development when undertaking the project. These include:

1. Research planning that aims to mitigate likely power imbalances between the project team and their participants;
2. Data analysis frameworks that help identify unique systemic complexities relating to agency, and a process of data synthesis that enables prioritization of areas for design intervention;
3. Co-design practices that both redistribute power between the team and participants and result in effective prototypes; and
4. A client engagement approach that positions them as partners so as to become advocates for the project.

The article comes to a close with final reflections about Navigating Home and the subsequent proposal that PSI design work requires a ‘reflexive practice’. This is a project approach that demands self-awareness in light of the literacy required to negotiate power and agency.

While aiming to describe best practices, this article also intends to share a candid portrayal of the challenges associated with doing work of this kind. It does so to explicitly address the dearth of realistic appraisals of service design work, particularly around issues of power. As Yoko Akama (2009) and others (Blomkamp 2018; Kimbell and Seidel 2008; Pierri 2017) have noted, service design-related case studies often lack critical reflection, in part due to non-disclosure agreements or a reluctance to air challenges that may dampen emerging interest in a new field. The outcome of authors offering only a starry-eyed view of their discipline is that newer practitioners and future clients do not understand how difficult and complex this work is, nor learn about risks and how they can be mitigated, as Pierri emphasizes: “serious analysis of agency and power are long overdue alongside more optimistic accounts of collaborative design projects, which are seen most of the time as ethical and good in their own right” (2017, S2952). This article has therefore been written for readers who may be: working in the social sector and service delivery; exploring issues of reflexivity and power in their work; interested in systems thinking and human-centered design; working with vulnerable populations; or looking to shift from commercial to social sector design work.

Adaptive Sampling

A project team can attempt to redistribute power by making project-level considerations about whom data is being collected about – in short, an approach to sampling that seeks fair representation. Project outcomes can only ever be as good as the quality of data informing them, therefore if the sample skews incorrectly it’s likely the project will suffer a correspondingly imbalanced outcome. When considered in the context of public sector work, design informed by poor data can result in project outcomes that at best, fail to adequately support the people using the service, and at worst, harm them. Yet, the recruitment of shelter residents can be challenging.

Following previous project work with shelter residents, the PPL team anticipated that recruitment and certainly long-term project participation – as is often desirable in human-centred design (and an aspiration in PSI design work) – would be made difficult by the transience that permeates residents’ lives. Residents can be hard to pin down: they drop in and out of the shelter system, rotate through shelter intakes and assessments, or are moved to new shelters with little notice. Many avoid their shelter except to sleep, instead keeping busy with work or study and a life beyond its walls. Institutionalizing shelter routines – such as curfews – act as a daily anchor, yet there is little that would enable the design team to secure reliable participation. Frontline staff can be similarly difficult to recruit. Their schedules change around residents’ needs and unanticipated events at the shelter. There could be no complete assurance that the team could engage a consistent sample of frontline staff or residents and fulfil specific demographic, let alone psychographic, sampling requirements.

Unsurprisingly, once in the field these concerns were realized. Not only was scheduling a consistent sample challenging, but on occasion shelters were simply too busy to recruit effectively for the team, as one PPL team member described: “sometimes they would grab who was nearby… sometimes they forgot that we were coming… that meant that sometimes the residents didn’t have the agency or the context of participating” (2019). The team were dismayed to hear one very anxious couple relay that they had been told that they were meeting someone from a government agency, rather than the independent design consultants patiently awaiting them. In these instances, the consent process (explained in detail below) became an essential tool for rebalancing power and allowing the resident to determine whether they even wanted to participate.

To counter inconsistent recruitment the team doubled down on data triangulation. If an interview didn’t yield the anticipated data, the team could at least gather a specific, highly detailed set of field observations, tap into the subject matter expertise of government agency and frontline staff, and seek out additional secondary research. In addition, they sought to hasten data analysis and synthesis by undertaking it while still in the midst of fieldwork (in the Burger King across the street, in cabs, the subway, while hunched over lunch during hurried breaks). This acceleration enabled the team to rapidly develop early hypotheses that could be tested early on, rather than risk discovering data gaps once back in the office after fieldwork was complete.

Humanizing Data Collection

While all research and design participants should feel that the terms of their contribution to a project are transparent and within their control, this is especially important when the participants are homeless shelter residents. Institutionalized living regularly undermines feelings of independence and self-actualization, meaning that residents may be more sensitive to perceptions that their agency is being further eroded. PPL employs a research ethics protocol for data collection and management that aims to elicit and honor how each individual participant wishes to share their data. This protocol is grounded by principles of causing no harm, being honest, and being fair.

At one level, this plays out in the emotional dexterity it takes to create a supportive conversation environment: being sensitive to participants’ emotional safety while also being judiciously helpful, directing residents to resources as needed, although never giving advice. Power imbalances are also addressed through specific data gathering and storage protocols that aim to give residents control over their privacy, such as through the way that data is captured and stored. For example, the consent form is the only document on which the participant’s name is recorded. All other materials or files are labelled with a unique participant code and scrubbed of sensitive data. At the conclusion of each interview participants are also offered the opportunity to change, delete or retract comments, as well as review and delete any photos taken of them.

Conduct-informed data collection is perhaps most explicitly expressed through PPL’s informed consent process. While gaining participant consent is standard in academic research and recognized as best practice in most industry-based work, PSI project teams need to be even more sensitive to and transparent with their participants. For Navigating Home, the interview moderator talked through each aspect of the consent language in aid of participants with low English literacy or disabilities that impede reading or written completion of the form. This was also a crucial opportunity to build rapport and trust in a bureaucratic moment that could otherwise easily alienate a population already subject to repetitive, and at times frustrating, paperwork. Like most consent forms, the project and discussion aims were stated in simple language, including who the project was for, how data would be used, and any risks or benefits to the participant. Yet, in light of the circumstances participants were living in, researchers also needed to be careful not to stoke false hopes by expressing unclear aspirations about the project’s impact, particularly in relation to the participant’s own circumstances or case.

PPL also aspires to give their participants granular control over the type of data being collected about them. For example, the participant had the option to circle ‘yes’ or ‘no’ to each form of data, including notetaking, audio-recording, video-recording, pictures in which their face couldn’t be seen, and pictures in which their face could be seen (see Figure 1). In addition, the consent form offered a list of resources relating to discrimination, the reporting of shelter issues, and mental health, which the participant could keep. While most residents responded to the carefully laid consent process with some version of “I’ve got nothing to hide”, it wasn’t uncommon for the team to have stories of depression, trauma or prejudice shared with them. Without supportive resources on hand the team simply would not have been able to exercise adequate duty of care while out in the field.

Figure 1. PPL’s consent form. Photograph ©Public Policy Lab, used with permission.

The team also came to understand that consent wasn’t simply a moment of interaction during the interview process. In one instance, as the research session progressed the moderator observed that the participant did not appear to be well and that there was a good chance that he would not have been able to fully comprehend the conditions of participation. While feeling conflicted by the risk of acting paternalistically, the moderator ultimately cut the interview short and withdrew the participant’s data from the project.

Fieldwork: Understanding End Users and Systems

The team undertook its first round of shelter fieldwork over the course of three blisteringly cold, mid-winter weeks. They travelled the furthest reaches of NYC’s ailing subway system, visiting twelve shelters for about three hours each. They aimed to speak with two residents and two frontline staff (usually a housing specialist and a case manager) per shelter, as well as hold shorter conversations with operational leadership, such as the shelter or program director. In total they met with twenty-six frontline staff and eighteen residents. As fieldwork progressed, each shelter began to feel like its own laboratory – host to its own dynamics, practices, stories, and historical trajectory. Some were excellent partners, mines for learning and design ideas, and offering a strong spirit of cooperation and even transparency about the challenges they were facing. Others were less capable of partnership, often due to structural destabilization caused by a transition in leadership. In these circumstances the team could not breach their outsider status and had to contend with reading between the lines of the view they were being offered by the shelter.

Defining Systemic Barriers

For Navigating Home, given the team had been tasked with identifying the system-wide factors that differentiate high- and low-performing shelters, the first step involved surfacing all of the factors affecting shelter performance, then distinguishing those that were beyond the scope of the project to affect. Data from participant observation alone was highly suggestive, in that it seemed to differentiate between low- and high-performing shelters in very visceral ways.

For example, the experience of arriving at some shelters was telling, with stringent security measures casting a foreboding welcome and residents listlessly clustering around building entrances. These were generally people who were unable to or couldn’t find work. They told the team that they’d usually be out in the neighborhood, but the bitter cold was preventing them from venturing further afield. In these shelters, staff appeared to work autonomously and within well-worn hierarchies.

However, other shelters were bustling, warm, with residents and staff trading jokes and banter. In these shelters, walls heaved with community notices, as well as shelter and social service information. There were DIY efforts, such as ‘welcome kits’ (pots, pans, towels) on display that residents would receive upon departure to their new home (see Figure 4), and even staff-made posters that visualized the pathway to rehousing. Some had computer labs and well-publicized life- or career-skill workshops. There were also those that were actively partnering with local churches or community organizations to gather donations, such as books and clothing, or even to run programming. During one shelter visit, the team’s interview was interrupted by the clanging of an old-fashioned handbell. Staff and residents filled the nearby corridor with an excited, congratulatory buzz – someone had found a home, and it seemed that everyone was celebrating.

Figure 4. A ‘welcome kit’ on display at a shelter. Photograph ©Public Policy Lab, used with permission.

Yet regardless of shelter environment, there were factors that seemed to affect all shelters and impede residents’ ability to find housing. These included the lack of affordable housing in NYC’s notoriously tight rental market, clients with profound needs (usually mental and physical illness or substance abuse), as well as covert – and illegal – landlord discrimination, in which prospective tenants described being rejected without clear reasoning, but suspected factors such as their current housing status, race, and family make-up.

Identifying Intervention Areas

Given all shelters were grappling with the systemic constraints described above, when identifying opportunities for intervention, further analysis found that difference in shelter performance often lay in the culture of service provision – of leadership, of staff collaboration, of the way each staff member chooses to conduct themselves. These differences manifest in concrete terms through shelter processes and outcomes, such as how proactive each shelter’s programming was, how effectively staff shared knowledge and best practices, the quality of tools and materials that staff were using to help residents, and how supported residents felt.

For example, the function of housing specialists – an explicit project focus, as described earlier – became a litmus for understanding how service experience might vary between low- and high-performing shelters. In higher-performing shelters, housing specialists were often positioned within a well-integrated, cross-functional team. Most frontline staff were equipped with enough housing knowledge to field residents’ basic questions, and housing specialists were sensitive to case management needs beyond housing. Program directors regularly gathered frontline staff to triage difficult cases and collaboratively develop a strategy to help residents progress. There were also practices in place for knowledge- and skills-sharing, so that everyone could be kept in the loop with informal – but important – information that the case management program doesn’t capture (e.g. a sick child). By contrast, staff at lower-performing shelters tended to work in isolation, reactively, were not supported as effectively, nor did they seem to be supporting each other. When considering these observations in response to the project goal of better leveraging housing specialists, it became clear that Navigating Home would need to develop design concepts that encouraged well-integrated, proactive staffing environments.

The team synthesized their data by collating the staff and resident needs relating to service experience that they had documented, then grouped them into sets of ‘shared needs’. These are needs that are jointly shared by staff and residents and indicative of service gaps, which, in short, are opportunities for design intervention. This practice of designing for shared needs aims to make a larger systemic impact by encouraging joint behavior change, therefore resulting in a more salient service intervention as it encourages positive behavioural interdependency between staff and residents. These shared needs were reframed as ‘service factors’, which, when well-managed, corresponded to higher shelter performance – as is described briefly below.

Designing for Service Factors

Trauma-Informed Co-Design

While PPL’s team has had experience working with at-risk populations, and some members had previously worked with homeless participants, none of the team are trained social workers or therapists. Some team members felt this gap acutely while conducting research in the shelters, grappling with the emotional challenge of bearing witness to trauma while simultaneously navigating the discomfort of their own comparative privilege. In addition, some struggled with feelings of powerlessness as they watched evidence mount about the magnitude of the factors that underlie homelessness – poverty, intergenerational racism, inequity – the systemic factors that lay beyond the scope of the project. As one team member reflected, it was confronting to witness how significantly systemic barriers were impacting people’s lives while also realizing the limits of one’s own agency as a practitioner, and to proceed knowing that no single project could resolve those intractable challenges.

“Initially it was very challenging because you are trying to get particular types of information from people and they are sharing other pieces of information that they feel comfortable sharing, but you know you have no way to help. They’re not even asking you to help, but even just as a caring human being – hearing that is tough. It’s actually emotionally draining… I was physically tired…. I felt bad that my help was not impactful. How do you navigate these spaces?” (PPL Team Member).

In recognition of the realization that a responsible co-design practice would need to both support the well-being of participants as well as the team itself, PPL enlisted the support of a clinical psychologist specializing in trauma. He visited the office bi-monthly for the remainder of the project, something of an experiment for both the therapist and PPL. During debrief sessions he guided the team through the trauma they were witnessing, while also exploring their feelings of frustration and disempowerment in not being able to have a more immediate impact. The team was coached on creating a reassuring and safe environment through conversation cues, as well as interview closure practices, such as inviting residents to linger with the team after the interview, should they need time to settle before departing the privacy of the interview room. The sessions thus became an important space for pragmatic reflection outside the rush of project work, marking a new process of periodic reflexivity that enabled the team to develop a practice of exercising improved duty of care with residents, themselves, and each other. In the interest of deepening trauma-informed practices and honouring a space of reflection, the psychotherapist meetings have become a permanent and valued fixture at PPL.

There’s a thing about boundaries. It’s about the emotional capacity and maturity that it takes to be an individual representing an organization while looking at another individual who is living in some incredibly difficult circumstances. To be in that moment, to actually be compassionate, to actually care about the individual, but know that you can’t actually change that individual’s life… The work we do is hard. I think it’s pretty foolish for any one of us to think that we could engage with some kind of difficulty in people’s lives and complex systems without observing the trauma affecting us… If you are drawn to this work it is because you care about the world … But you’ve also made a strategic choice to do that at a systems level… and complicated feelings are going to come up with that (PPL Director 2019).

Outputs: Service Concepts

For the co-design phase the team returned to a selection of their original shelter sample, aiming for a tighter, more focused engagement with a smaller cohort of participants. Where possible they invited participation from their first participant sample, given that a trusted relationship had already been established. In total, they conducted co-design sessions with twenty-nine frontline staff and twenty-three residents. Putting trauma-informed co-design practices into play, these residents and staff members guided the team through the process of identifying prototype use cases, turning rough single-page sketches into journey maps and worksheets, parsing language into more relatable terms, and helping a visual language emerge (see Figure 6).

Figure 6. A frontline staff member participating in a co-design activity with PPL. Photograph ©Public Policy Lab, used with permission.

The team also took advantage of an opportunity to hold co-design sessions within the agency’s quarterly housing specialist trainings, which every one of NYC’s housing specialists was due to attend. They ran an activity that encouraged staff to design their ‘dream training’ with respect to content and format, and a mapping exercise to identify residents’ needs when moving into permanent housing (see Figure 7).

Figure 7. Housing specialists participating in a co-design session with PPL. Photograph ©Public Policy Lab, used with permission.

The final outcomes of co-design were four service concepts spanning a range of intervention areas. The service concepts were validated through prototype-testing in which new tools, materials, and processes were designed with residents and staff, and that aim to encourage behaviour change in interactions between staff and residents, as well as support intrinsic motivation and informed decision-making. The four service concepts, along with the design objective guiding their direction, are described at a high level below.

Participatory Partnership

When working with government agencies PPL aims to position itself as a facilitator and enabler. They take measures that encourage their clients and project stakeholders to take ownership of the project and become enthusiasts for its implementation. This approach is driven by a belief that the government agency is not the ultimate client needing to be served, as described by a PPL director: “We don’t call government agencies our clients. Government agencies are our partners, and we have a mutual client, which is the public. What we’re all working towards is increased opportunity for members of the public” (2019).

One learned best practice involves engaging an ‘agency leadership from the project outset. This group comprises executive and programmatic leaders who hold operational decision-making authority. As the future life of the project rests on project stakeholders’ shoulders, it’s important to cultivate ‘authentic trust’ so that the working relationship is sincere, reflective, and pragmatic (Solomon and Flores 2001). In the project plan, PPL outlines collaboration requirements to this end, including a request for a number of project partners to embed in the project team, attend fieldwork, and participate in internal work sessions. This invitation is similarly extended to the project funders. For Navigating Home PPL also worked with their primary government point-of contact to assemble a twenty- to thirty-person agency leadership group comprising staff who specialize in areas such as shelter operations, specific resident populations, staff training, and legal affairs. The group was kept abreast of project developments and attended ‘pin-up’ meetings following each project milestone, in which they contributed feedback and expertise to the project, as well as participated in co-design activities. These activities were essential for balancing user experience with operationally feasibility, and to help craft short- and long-term goals for implementation. Without this collaborative engagement the project team risked developing design work that lacked the necessary buy-in and operational backbone for successful implementation. Following project implementation PPL will also conduct a series of progress evaluation check-ins at the three-, six-, and twelve-month marks, so as to support their partners with arising challenges and the task of assessing project impact.

Reporting With Agency Coordination Tools

Throughout Navigating Home, reporting wasn’t a moment, but a process. Every meeting and pin-up contributed to the cumulative effort of equipping government partners and stakeholders with the stories, materials, and knowledge they needed to become advocates for the project and facilitate its implementation. This was a process of transferring agency through evidence-based storytelling that encouraged empathy for the people being served by the project, while also offering a pragmatic operational path forward, with the right level of strategic detail.

As the team regrouped after their penultimate pin-up, they noted that their stakeholder group and agency partners had focused on individual prototypes as catalysts for change. This was unsurprising, given that the team had been sharing discrete service interactions and their related prototypes in pin-ups, as compared to demonstrating distributed impact through the range of end-to-end service interactions they had been designing. Designing a service is ultimately about curating a flow of interactions, rather than a series of artifacts, as described by Akama: “designers are defined by what they enable, not what they ‘make’” (2009, 7). The team realized that they would need to better represent the deliverables at a systemic level, highlighting time, space, and human interaction, if to successfully convey the impact Navigating Home could make. The team therefore designed an ‘interaction model’ – a visual artifact that demonstrated resident and frontline staff interactions through one continuous, systemic experience, rather than a one-off interaction with a tool. While the tools they had designed could indeed be used independently, the team wanted to reflect fieldwork findings about client-centered, holistic, team-based models for work that suggested the tools would be most impactful when implemented together to support collaborative work practices. In their final presentation they breathed life into the interaction model by presenting it through narrative-based storytelling.

The team supplemented the system view with a set of policy and operational recommendations to address the issues that couldn’t be designed for within the scope of Navigating Home, but would nonetheless impact move-out rates immensely, if they could be addressed. These recommendations covered factors relating to staff training, suggestions for staffing and resourcing, conducting assessments, and both inter- and intra-agency workflow. These recommendations were shared with the belief that systemic impact can be greater when changes are made further up the power chain. Rather than frontline staff and residents largely shouldering the burden of responsibility for improved service outcomes, upstream policy changes could have significant flow-on effects: if programmatic agency staff are able to work more effectively within improved policy conditions, then they can better support frontline staff in their task of supporting residents.