Governance structures for new technologies are frequently top-down, reactive, and informally enforced, leaving marginalized communities with little power to address harms until after they occur. To address these limitations, we introduce Proactive Accountability, a conceptual framework theorizing that effective governance must be community-led, formally enforced, and continually maintained. We explore these principles through a speculative design study with Detroit’s food sovereignty community, in which participants identified community-owned cooperatives—described as “ancestral technologies”—as a model for redistributing power within a capitalist economy. Synthesizing these theoretical and empirical insights, we introduce the Designing for Proactive Accountability (D4PA) framework, providing implications for how designers can operationalize the goals of proactive accountability into HCI research and design projects. Finally, we contribute a future research agenda that positions cooperatives not merely as beneficiaries of design, but as sites of inquiry for understanding how to institutionalize justice-oriented democratic governance of sociotechnical systems.

Proponents of legal automation believe that translating the law into code can improve the legal system. However, research and reporting suggest that legal software systems often contain flawed translations of the law, resulting in serious harms such as terminating children's healthcare and charging innocent people with fraud. Efforts to identify and contest these mistranslations after they arise treat the symptoms of the problem, but fail to prevent them from emerging. Meanwhile, existing recommendations to improve the development of legal software remain untested, as there is little empirical evidence about the translation process itself. In this paper, we investigate the behavior of fifteen teams---nine composed of only computer scientists and six of computer scientists and legal experts---as they attempt to translate a bankruptcy statute into software. Through an interpretative qualitative analysis, we characterize a significant epistemic divide between computer science and law and demonstrate that this divide contributes to errors, misunderstandings, and policy distortions in the development of legal software. Even when development teams included legal experts, communication breakdowns meant that the resulting tools predominantly presented incorrect legal advice and adopted inappropriately harsh legal standards. Study participants did not recognize the errors in the tools they created. We encourage policymakers and researchers to approach legal software with greater skepticism, as the disciplinary divide between computer science and law creates an endemic source of error and mistranslation in the production of legal software.

As algorithms become an influential component of government decision-making around the world, policymakers have debated how governments can attain the benefits of algorithms while preventing the harms of algorithms. One mechanism that has become a centerpiece of global efforts to regulate government algorithms is to require human oversight of algorithmic decisions. Despite the widespread turn to human oversight, these policies rest on an uninterrogated assumption: that people are able to effectively oversee algorithmic decision-making. In this article, I survey 41 policies that prescribe human oversight of government algorithms and find that they suffer from two significant flaws. First, evidence suggests that people are unable to perform the desired oversight functions. Second, as a result of the first flaw, human oversight policies legitimize government uses of faulty and controversial algorithms without addressing the fundamental issues with these tools. Thus, rather than protect against the potential harms of algorithmic decision-making in government, human oversight policies provide a false sense of security in adopting algorithms and enable vendors and agencies to shirk accountability for algorithmic harms. In light of these flaws, I propose a shift from human oversight to institutional oversight as the central mechanism for regulating government algorithms. This institutional approach operates in two stages. First, agencies must justify that it is appropriate to incorporate an algorithm into decision-making and that any proposed forms of human oversight are supported by empirical evidence. Second, these justifications must receive democratic review and approval before the agency can adopt the algorithm.

Governments are increasingly turning to algorithmic risk assessments when making important decisions, such as whether to release criminal defendants before trial. Policymakers assert that providing public servants with algorithmic advice will improve human risk predictions and thereby lead to better (e.g., fairer) decisions. Yet because many policy decisions require balancing risk-reduction with competing goals, improving the accuracy of predictions may not necessarily improve the quality of decisions. If risk assessments make people more attentive to reducing risk at the expense of other values, these algorithms would diminish the implementation of public policy even as they lead to more accurate predictions. Through an experiment with 2,140 lay participants simulating two high-stakes government contexts, we provide the first direct evidence that risk assessments can systematically alter how people factor risk into their decisions. These shifts counteracted the potential benefits of improved prediction accuracy. In the pretrial setting of our experiment, the risk assessment made participants more sensitive to increases in perceived risk; this shift increased the racial disparity in pretrial detention by 1.9%. In the government loans setting of our experiment, the risk assessment made participants more risk-averse; this shift reduced government aid by 8.3%. These results demonstrate the potential limits and harms of attempts to improve public policy by incorporating predictive algorithms into multifaceted policy decisions. If these observed behaviors occur in practice, presenting risk assessments to public servants would generate unexpected and unjust shifts in public policy without being subject to democratic deliberation or oversight.

We introduce a new framework for conceiving of and studying algorithms that are deployed to aid human decision making: “algorithm-in-the-loop” systems. The algorithm-in-the-loop framework centers human decision making, providing a more precise lens for studying the social impacts of algorithmic decision making aids. We report on two experiments that evaluate algorithm-in-the-loop decision making and find significant limits to these systems.

The rise of machine learning has fundamentally altered decision making: rather than being made solely by people, many important decisions are now made through an "algorithm-in-the-loop'' process where machine learning models inform people. Yet insufficient research has considered how the interactions between people and models actually influence human decisions. Society lacks both clear normative principles regarding how people should collaborate with algorithms as well as robust empirical evidence about how people do collaborate with algorithms. Given research suggesting that people struggle to interpret machine learning models and to incorporate them into their decisions---sometimes leading these models to produce unexpected outcomes---it is essential to consider how different ways of presenting models and structuring human-algorithm interactions affect the quality and type of decisions made. This paper contributes to such research in two ways. First, we posited three principles as essential to ethical and responsible algorithm-in-the-loop decision making. Second, through a controlled experimental study on Amazon Mechanical Turk, we evaluated whether people satisfy these principles when making predictions with the aid of a risk assessment. We studied human predictions in two contexts (pretrial release and financial lending) and under several conditions for risk assessment presentation and structure. Although these conditions did influence participant behaviors and in some cases improved performance, only one desideratum was consistently satisfied. Under all conditions, our study participants 1) were unable to effectively evaluate the accuracy of their own or the risk assessment's predictions, 2) did not calibrate their reliance on the risk assessment based on the risk assessment's performance, and 3) exhibited bias in their interactions with the risk assessment. These results highlight the urgent need to expand our analyses of algorithmic decision making aids beyond evaluating the models themselves to investigating the full sociotechnical contexts in which people and algorithms interact.

After decades of urban investment dominated by sprawl and outward growth, municipal governments in the United States are responsible for the upkeep of urban neighborhoods that have not received sufficient resources or maintenance in many years. One of city governments' biggest challenges is to revitalize decaying neighborhoods given only limited resources. In this paper, we apply data science techniques to administrative data to help the City of Memphis, Tennessee improve distressed neighborhoods. We develop new methods to efficiently identify homes in need of rehabilitation and to predict the impacts of potential investments on neighborhoods. Our analyses allow Memphis to design neighborhood-improvement strategies that generate greater impacts on communities. Since our work uses data that most US cities already collect, our models and methods are highly portable and inexpensive to implement. We also discuss the challenges we encountered while analyzing government data and deploying our tools, and highlight important steps to improve future data-driven efforts in urban policy.