The AI Incident Database as an Educational Tool to Raise Awareness of AI Harms: A Classroom Exploration of Efficacy, Limitations, & Future Improvements
Project Overview
The document explores the integration of the AI Incident Database (AIID) as an innovative educational resource designed to raise awareness about AI-related harms in computer science education. It highlights a classroom activity that utilizes AIID, employing pre- and post-activity questionnaires to evaluate shifts in students' understanding and attitudes toward AI ethics, safety, and accountability. The findings indicate that interaction with AIID significantly enhances students' comprehension of the prevalence and severity of AI-related issues. However, the study also points out that while AIID serves as a valuable tool for fostering ethical awareness, there are areas for improvement in its usability and the clarity of its reporting mechanisms. Overall, the document underscores the potential of generative AI tools like AIID to enrich educational experiences by promoting critical discussions around the ethical implications of AI technologies.
Key Applications
AI Incident Database (AIID)
Context: Graduate-level computer science course focusing on societal and ethical implications of AI and ML
Implementation: Classroom activity involving pre-activity questionnaire, interaction with AIID, post-activity questionnaire, and class discussion
Outcomes: Increased awareness of AI harms, shift in values towards prioritizing safety and accountability, enhanced motivation for the course
Challenges: Limited to a small sample size and single institution, potential biases in participant selection, usability issues with AIID
Implementation Barriers
Usability Barrier
Some students found the AI Incident Database difficult to use, with reports of it being hard or very hard to navigate.
Proposed Solutions: Recommendations for improving the user interface and enhancing the database's usability through automation and better promotion.
Generalizability Barrier
The findings are based on a small, homogenous sample of students from a single institution, limiting the ability to generalize results.
Proposed Solutions: Encouragement for future studies to replicate the research in diverse educational settings and with larger sample sizes.
Engagement Barrier
Students relied heavily on social media for learning about AI incidents rather than educational materials, which indicates a lack of engagement with formal educational content.
Proposed Solutions: Emphasizing the importance of integrating AI ethics into standard curricula to improve awareness and foster engagement.
Project Team
Michael Feffer
Researcher
Nikolas Martelaro
Researcher
Hoda Heidari
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Michael Feffer, Nikolas Martelaro, Hoda Heidari
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai