A Typology of Trust for AI: Design An Intervention to A Specific AI Harm
There is an extensive international conversation around the how to govern the use artificial intelligence. This attention is warranted, as large-scale algorithmic decision-making systems are proliferating across the many industries. Journalists and academics have identified a wide range of problems with these models (see below), but government has done little to protect citizens. Since, AI differs dramatically across its many uses, and no single regulatory intervention makes sense for all AI systems.
Work with a team to find a solution to problematic AI applications. First, choose a problematic instance of AI—you can use the list below or find your own problem. Next, work through possible interventions. Which would work well to alleviate the systemic AI issue?
Interventions: Interpretability, explainability; Protect tech whisteblowers; Algorithmic audits; Ex-ante approval; Public reporting; Problems: Justice systems; Discriminatory healthcare/employment algorithms; College admissions/financing; Credit approval; Critical infrastructure; Safety of autonomous cars;