Skip to main content
SearchLoginLogin or Signup

Empowered Participatory Design in Algorithm Design for the U.S.Child-Welfare System

Contribution to the PDC 2020 Interactive Workshop "Computing Professionals for Social Responsibility: The Past, Present and Future Values of Participatory Design".

Published onMay 14, 2020
Empowered Participatory Design in Algorithm Design for the U.S.Child-Welfare System

Empowered Participatory Design in Algorithm Design for the U.S.Child-Welfare System

By Devansh Saxena, Shion Guha

Department of Computer Science, Marquette University Milwaukee, Wisconsin, USA

Contact: {devansh.saxena, shion.guha}

The Child-Welfare System (CWS) in the United States has come under escalating public and media scrutiny because of the potential damage done to children who are inappropriately removed from the care of their parents [4]. CWS has increasingly turned towards algorithms as a way of standardizing decisions and demonstrating that these decisions are unbiased and evidence-based [13]. Moreover, CWS in almost every state is underfunded and algorithmic systems offer a means to reduce costs [14]. Our research focuses on the collaborative work of child-welfare teams that participate in meetings mediated by policy, practice, and algorithms. The caseworkers’ job is to navigate through the system and find services for children and families as well as act as a mediator between the children’s court, district attorney’s office, and families. They are not trained to evaluate algorithmic decisions or think statistically, per se [7]. Caseworkers will often treat the decisions from these probabilistic models as perfect predictions and are more inclined to rely on these decisions over their own judgments that are based on experience and individual heuristics [17, 7]. Empowered participatory design (PD) can help address these issues by ensuring that caseworkers’ values are embedded into the design process through a mutual exchange of knowledge between the co-designers while concurrently acknowledging and addressing the power imbalance that exists in the public sector.

The democratization of data, easy access to software APIs, and neoliberal politics based in austerity and privatization has lead to the opaque and isolated development of sociotechnical systems with a decline in practices like participatory design (PD) and value-sensitive design (VSD) [10]. We are seeing a similar trend in the public sector where bills are being introduced in the Congress of the United States that endorse the development of predictive analytics in the child-welfare system [1]. Here, an organization like Computing Professionals for Social Responsibility (CPSR) can act as a data ethics committee that advises policymakers on decisions concerning complex sociotechnical systems. CPSR can provide a guiding framework that ensures PD and VSD are conducted ethically and under proper oversight.


Wisconsin’s Department of Children and Families (DCF) contracts child-welfare services to non-profit organizations (NGOs) that work in human services. We are conducting an ethnographic study with SaintA [15], one such NGO, and are investigating two algorithms that influence the work of child- welfare teams. One of these algorithms is a legislative requirement mandated by the state while the other is designed by the NGO itself to study progress around theory-driven practice based in trauma-informed care [19].

Child and Adolescent Needs and Strength (CANS) algorithm is mandated by the state and is used to assess the level of need of a foster child by determining the associated risk factors as well as well-being indicators. However, this algorithm often does not agree with theory-driven practice founded in trauma-informed care. Moreover, it has been re-appropriated to carry out a task that it was not designed to do which has led to several unintended consequences and added to the frustrations of caseworkers (see [16] for details).

Seven essential ingredients (7ei) algorithm is used to capture a child’s well-being over the course of the child-welfare case. The team scores the child’s wellness on seven categories (Prevalence, Impact, Perspective Shift, Regulation, Relationship, Reasons to Be, and Caregiver Capacity) every month when the case is discussed. 7ei algorithm is based in trauma-informed care which has been proven to improve child outcomes such as placement stability and permanence [19]. There are some concerns about the reliability of some measures but the program directors are aware of this and are working with their teams to resolve these issues.


The design process in PD is just as important as the end result [20]. We advocate for a design process that is conducted under proper oversight such that it recognizes asymmetrical power relations and vigorously seeks the inclusion of marginalized voices. An empowered design process can help PD in child-welfare in the following ways -

Mutual Learning - Introducing policy-mandated algorithmic systems in child-welfare means that caseworkers must continue to use a resource that is beyond their area of expertise even when it conflicts with their contextual judgments [16, 18]. This is where the design process is an integral part of PD where the exchange of knowledge must occur between the co-designers [20]. The design process allows researchers to learn about the caseworkers’ practice as well as their values. Understanding how caseworkers’ frame children in terms of trauma-informed care will help design algorithms that reflect these values and guide their intended use. In addition, caseworkers who are not traditionally trained in evaluating algorithms get to learn about the inner workings of the algorithm, thereby, unraveling the black-box nature while augmenting trust and explainability [3].

Exchange of values - Majority of the decision-making power sits with legal parties (district attor- ney’s office and judges) with caseworkers finding themselves at the periphery of the judicial process and lacking agency. Their values are marginalized because the sociological understanding of "child safety" is superseded by a legal understanding. PD employs value practices that can help address these differing meanings of inherently the same values placed in child-safety. PD allows for "communi- cation across irreducible differences" [8] such that the design space can become a space for pluralism of values where an agnostic and non-coercive consensus can be reached [12]. Here, academic researchers can act as mediators in the process where the stakeholders are able to re-negotiate their positions [9]. Moreover, PD’s wide variety of tools and techniques can be used to explore this value pluralism while suspending any immediate decisions with respect to design formulation [2]. Furthermore, embedding the values of theory-based practice (in this case, trauma-informed care [19]) into design ensures that practice is the driving force behind decision-making processes.

Equalising power relations - In a complicated domain like child-welfare, phronetic research [20, 6] has an important role to play in that the researchers can provide detailed narratives of power relations and asymmetries, and how power works and to what consequences [5]. This is especially important in public services where reducing costs always appears to be the primary motive. Unraveling these power dynamics and offering transparency to the public allows for activists-lead reform such that the motives are refocused towards the well-being of children [11]. This can in turn lead to the re- centering of design in the caseworkers’ rights and values who directly interact with children and families, advocate for them, and are responsible for their well-being.


We discuss some of the barriers we have experienced in our efforts to conduct participatory design –

Legal barriers - We experienced a deep commitment and advocacy towards families from social workers at the agency level who welcomed our presence as researchers. Indeed, when we started an interview study, some caseworkers (including a program director) reached out to us to be included in the study. This active involvement from the agency stakeholders will allow us to help them revise the 7ei algorithm using the empowered design process that we mapped out. However, at the state level, we have experienced several legislative barriers while working with the state to complete a data- sharing agreement. This process has been underway for more than 18 months with several revisions of the agreement submitted by us and reviewed by lawyers, and yet more revisions requested. Here, an organization like CPSR can help re-negotiate positions between researchers and political actors and help build mutual trust and accountability.

High turnover – CWS experiences a high turnover rate and it poses a barrier to a continued commitment towards participatory design. Indeed, two of the three NGO personnel with whom we started this research collaboration have already moved on to positions with different organizations. Therefore, it is imperative to continue to build relationships with both the new and seasoned employees at such organizations. An organization like CPSR can help foster a continued relationship and a mutual commitment between academic researchers and NGOs in human services.

Different contractors – The state uses a centralized data platform (which is utilized by all the NGOs) which contains all the information pertinent to child welfare cases. However, the state con-

tracts the development of different features within the platform to different companies. This is a complicated ecosystem that is composed of several NGOs and contractors and poses a barrier to con- ducting participatory design. Each NGO also maintains its own data platform and therein lies some hope for participatory design. Nevertheless, it is hard to gain collective buy-in beyond the agency level and affect change at the state level. Here, CPSR can help negotiate positions with political actors and act as an oversight committee for these contractors and ensure that caseworkers’ voices are not marginalized in the process.


  1. Using Data To Help Protect Children and Families Act. 115th Congress, Senate of the United States, 2018.

  2. Tone Bratteteig and Ina Wagner. Disentangling participation: power and decision-making in participatory design. Springer, 2014.

  3. Florian Cech. Beyond transparency: Exploring algorithmic accountability. In Companion of the 2020 ACM International Conference on Supporting Group Work, pages 11–14, 2020.

  4. Lindsay D Cooper. Implications of media scrutiny for a child protection agency. J. Soc. & Soc. Welfare, 32:107, 2005.

  5. Pelle Ehn and Richard Badham. Participatory design and the collective designer. In Proceedings of the Partici- patory Design Conference, pages 1–10. Malmö, Sweden, 2002.

  6. Bent Flyvbjerg. Making social science matter: Why social inquiry fails and how it can succeed again. Cambridge university press, 2001.

  7. Eileen Gambrill and Aron Shlonsky. Risk assessment in context, 2000.

  8. Donna Haraway. Situated knowledges: The science question in feminism and the privilege of partial perspective.

Turning points in qualitative research: Tying knots in a handkerchief, 2003:21–46, 2003.

  1. Ole Sejer Iversen, Kim Halskov, and Tuck W Leong. Values-led participatory design. CoDesign, 8(2-3):87–103, 2012.

  2. Michael Kaethler, Seppe De Blust, and Tim Devos. Ambiguity as agency: critical opportunists in the neoliberal city. CoDesign, 13(3):175–186, 2017.

  3. Anne Marie Kanstrup and Ellen Christiansen. Selecting and evoking innovators: combining democracy and creativity. In Proceedings of the 4th Nordic conference on Human-computer interaction: changing roles, pages 321–330, 2006.

  4. Chantal Mouffe. The return of the political, volume 8. Verso, 2005.

  5. Kathleen G Noonan, Charles F Sabel, and William H Simon. Legal accountability in the service-based welfare state: Lessons from child welfare reform. Law & Social Inquiry, 34(3):523–568, 2009.

  6. Jeanne S Ringel, Dana Schultz, Joshua Mendelsohn, Stephanie Brooks Holliday, Katharine Sieck, Ifeanyi Edochie, and Lauren Davis. Improving child welfare outcomes: balancing investments in prevention and treatment. Rand health quarterly, 7(4), 2018.

  7. SaintA. Trauma informed care: Seven essential ingredients, 2020., Last accessed on 2020-5-12.

  8. Devansh Saxena, Karla Badillo-Urquiola, Pamela Wisniewski, and Shion Guha. Child welfare system: Interaction of policy, practice and algorithms. In Companion of the 2020 ACM International Conference on Supporting Group Work, pages 119–122, 2020.

  9. Devansh Saxena, Karla Badillo-Urquiola, Pamela J Wisniewski, and Shion Guha. A human-centered review of algorithms used within the us child welfare system. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pages 1–15, 2020.

  10. Craig S Schwalbe. Strengthening the integration of actuarial risk assessment with clinical judgment in an evidence based practice framework. Children and Youth Services Review, 30(12):1458–1464, 2008.

  11. James Topitzes, Timothy Grove, Erika E Meyer, Stacey M Pangratz, and Caitlin M Sprague. Trauma-responsive child welfare services: A mixed methods study assessing safety, stability, and permanency. Journal of Child Custody, 16(3):291–312, 2019.

  12. Maja Van der Velden, C Mörtberg, J Van den Hoven, PE Vermaas, and I Van de Poel. Participatory design and design for values. Development, 11(3):215–236, 2014.

No comments here
Why not start the discussion?