Send message to Contributor

Short abstract:

It is a short activity that's a cross between an interactive presentation and an installation.

Abstract:

As regulation to hold AI accountable picks up steam with the advent of the EU AI Act, we reimagine policy construction as a community learning process rather than a top-down prescription from authorities. This activity invites conference attendees to assemble their idealAI accountability policy by combining snippets from a pool of potential printed options. Participants will select a regulatory scenario among domains of finance, medicine, or recommender-systems and put together text that aligns with the values they want to prioritize for policy. They will physically organize their policy by rearranging fabric pieces with individual requirements onto a corkboard. The policy text samples are drawn from a collaborative research exercise in which a sociology and computer science undergraduates simultaneously designed AI systems and policies governing explanations for inscrutable AI systems. This exercise invites conference attendees to confront some of the same tradeoffs that we found in our research on explainability for AI systems, like the balance between comprehensiveness and user understandability, and extend our research into further possibilities. We will pin together each attendee-generated policy on a corkboard and take a photograph. These policies will subsequently be uploaded to a website that will feature a digital version of the activity to gather input from the general public. Additionally, users will be able to view and comment on the configurations of other AI policies.