In the past few years, we have witnessed both a stunning acceleration of artificial intelligence (AI) innovation and a suite of terrifying AI blunders. From racist predictive sentencing software to sexist automated resume processing, and from recommendation algorithms that radicalize us to feeds that distract us, it is clear that AI has had more than its fair share of unintended consequences.
This website documents our 2023-2024 course in the Honors College at the University of Utah, which investigated current AI failures in the domains of bias and discrimination, radicalization and polarization, privacy and surveillance, environmental sustainability, and labor. We engaged deeply with questions about the ethics, unintended consequences, and potential futures of AI through readings, discussions, case-studies, guest lectures, playing with code, and field trips. We asked what kinds of processes, questions, and safeguards could have prevented the worst consequences of prominent AI mistakes and also sought inspiration from ongoing efforts to encourage more responsible AI.
Learn
Fall 2023 Semester.
- Read about prominent AI failures
- Hear from leading experts on AI
- Discuss ways to mitigate AI harms
- Audit public-facing AI systems in preparation for larger class audit
Research
Spring 2024 Semester.
- Choose socially-relevant AI: Pre-trial Sentencing Algorithms in Utah
- Unpack PSA algorithm and context
- Design and publish AI intervention
Transform
Course Outcomes.
- History and timeline of PSA deployment in Utah
- PSA Simulation + Tabletop Game on AI
- Video essay on PSA
- Collection of AI Audits
- Letters to legislators and editors
The goal of this course was to equip students with the sociotechnical and critical skills to think through the human consequences of decisions made in the creation and deployment of AI. Now we are employing these critical and technical skills in a real-world projects like AI audits and the investigation of a particular piece of software in Utah, the PSA risk assessment tool that is used in pre-trial detention decisions.
We are interested in the story behind how an algorithm was adopted in Utah because we know that our state will be pitched AI to solve problems in the future, and we want to investigate how those decisions can be more informed, democratic, and with a robust evaluation plan in place.
Instructors:
Elizabeth Callaway, Department of English
Rogelio E. Cardona-Rivera, Division of Games
University of Utah