Funder Update Prep Pack

Source-grounded context summary

The project explores a scalable academic integrity model for the generative AI era. It uses game theory and examples from sampling-based enforcement systems to propose random learning audits as a potentially less invasive alternative to proctoring software.

Two-minute update script

This working paper reframes AI-enabled cheating as an incentive problem. Instead of asking institutions to detect every use of AI, it asks whether assessment systems can make cheating a bad gamble by requiring some students to explain their submitted work after the fact. The model suggests that credible random audits, meaningful consequences, and strong fairness safeguards may deter cheating with a manageable amount of instructor time. The next step would be small pilot studies to test workload, student experience, accessibility, and deterrent effects.

Progress / value framing

• Offers a concrete, testable model rather than a general critique of proctoring.
• Connects academic integrity to established enforcement logic from other sectors.
• Identifies fairness and false-positive risk as central design constraints.

Likely funder questions

• What evidence would a pilot need to collect?
• How would the team protect students with anxiety, disabilities, or language barriers?
• What institutional approvals would be needed before testing?

Risks, limits, and verification notes

The current paper is model-based. Any funder-facing update should be clear that empirical validation remains future work.