Search for: All records

Despite the growing interest in human-AI decision making, experimental studies with domain experts remain rare, largely due to the complexity of working with domain experts and the challenges in setting up realistic experiments. In this work, we conduct an in-depth collaboration with radiologists in prostate cancer diagnosis based on MRI images. Building on existing tools for teaching prostate cancer diagnosis, we develop an interface and conduct two experiments to study how AI assistance and performance feedback shape the decision making of domain experts. In Study 1, clinicians were asked to provide an initial diagnosis (human), then view the AI's prediction, and subsequently finalize their decision (human-AI team). In Study 2 (after a memory wash-out period), the same participants first received aggregated performance statistics from Study 1, specifically their own performance, the AI's performance, and their human-AI team performance, and then directly viewed the AI's prediction before making their diagnosis (i.e., no independent initial diagnosis). These two workflows represent realistic ways that clinical AI tools might be used in practice, where the second study simulates a scenario where doctors can adjust their reliance and trust on AI based on prior performance feedback. Our findings show that, while human-AI teams consistently outperform humans alone, they still underperform the AI due to under-reliance, similar to prior studies with crowdworkers. Providing clinicians with performance feedback did not significantly improve the performance of human-AI teams, although showing AI decisions in advance nudges people to follow AI more. Meanwhile, we observe that the ensemble of human-AI teams can outperform AI alone, suggesting promising directions for human-AI collaboration. 

Government use of algorithmic decision-making (ADM) systems is widespread and diverse, and holding these increasingly high-impact, often opaque government algorithms accountable presents a number of challenges. Some European governments have launched registries of ADM systems used in public services, and some transparency initiatives exist for algorithms in specific areas of the United States government; however, the U.S. lacks an overarching registry that catalogs algorithms in use for public-service delivery throughout the government. This paper conducts an inductive thematic analysis of over 700 government ADM systems cataloged by the Algorithm Tips database in an effort to describe the various ways government algorithms might be understood and inform downstream uses of such an algorithmic catalog. We describe the challenge of government algorithm accountability, the Algorithm Tips database and method for conducting a thematic analysis, and the themes of topics and issues, levels of sophistication, interfaces, and utilities of U.S. government algorithms that emerge. Through these themes, we contribute several different descriptions of government algorithm use across the U.S. and at federal, state, and local levels which can inform stakeholders such as journalists, members of civil society, or government policymakers