/
Task Force Meeting 13
Edit

Task Force Meeting 13

Meeting Recording

Quick recap

The meeting focused on discussing the challenges and limitations of detecting AI-generated content in academic settings, particularly in teaching and research contexts. The group explored various AI detection tools and their effectiveness, including their false positive rates and inability to reliably distinguish between human-written and AI-generated text. They concluded by planning future activities including an AI Research Day and a lunch meeting with the new director of DSAI, while also discussing best practices for grant submissions and the importance of clear communication about AI policies.

Next steps

  • Junie and Gregory: Figure out institutional access to Proofig and other AI detectors through institutional licenses

  • Kadija: Email David Dowdy about creating a SharePoint resource area or conference for sharing AI strategies in public health education, and CC Gregory

  • PHAISE working group: Generate best practices document for using AI detectors based on this discussion, to be shared with everyone and posted on wiki site

  • Junie and Gregory: Update AI-related materials on wiki site periodically and use PHAISE meetings to review for outdated content

  • Junie: Share AI Research Day call for abstracts and submission form documents

  • Gregory and Junie: Send out call for AI Research Day applications in early January

  • Interested PHAISE members: Reach out to Junie and Gregory if available for January 14th lunch with Mark Drezde

  • Junie and Gregory: Coordinate with Keisha/Data Council regarding January 14th lunch slots and figure out overlap with existing

Ethical AI in Education Challenges

Brian discussed his teaching experience with generative AI and public health, emphasizing the importance of ethical usage and awareness of AI's limitations. He explained the challenges of detecting AI-generated content, highlighting the arms race between AI users and detectors. Brian also shared a practical example of a faculty member using a simple method to catch students using AI for assignments.

AI Detection Tool Limitations

Brian discussed the limitations and inaccuracies of AI detection tools, highlighting their high false positive rates and inability to reliably distinguish between human-written and AI-generated text. He shared examples demonstrating how these tools often fail to detect AI-generated content, especially in non-native English writing and when used for light editing. Gregory noted that Proofig is commonly used in basic science research for detecting AI-generated images, and the group discussed the need to verify institutional access to such tools. Brian recommended avoiding AI detection tools altogether and suggested citing relevant studies to defend against accusations of AI use if necessary.

AI Detection in Research Submissions

The group discussed the use of AI detection tools for research submissions, particularly focusing on NIH guidelines and plagiarism concerns. Gregory explained that while there is no specific implementation guidance from NIH, the requirement for original submissions is intended to prevent excessive AI-generated work, citing an example of a faculty member submitting 47 grants in a calendar year. The discussion highlighted that while tools like Proofig are commonly used in basic science to detect figure manipulation, their broader use in grant submissions remains unclear, though Gregory suggested that AI detection might become an expected practice, especially in cases of research misconduct.

AI Grant Writing Best Practices

The group discussed AI detection and best practices for grant submissions, with Brian emphasizing the importance of not overstressing faculty and encouraging clear communication about AI policies. They agreed to compile materials on their wiki site, including AI disclosure statements and guidelines for using AI in grant writing. The group also planned AI Research Day for March 23rd, which will include a keynote speaker, poster presentations, and potential awards. Finally, they discussed a lunch meeting with Mark Drezde, the new director of DSAI, scheduled for January 14th, 2026.