Rethinking Brown Economics AI Cheating

I have been thinking a lot about the recent story of rampant AI-based cheating in a welfare economics class at Brown. This Inside Higher Ed story captures the story.

I might be wrong, but I think most of these stories are framing the problem backward. Instead of framing it as how do we stop students from cheating with AI. We should be changing the way we teach to encourage, if not require, the use of AI, but have evaluations (exams, tests, etc.) require a level of critical thinking that AI cannot meet.

Over the last 100 years, education has had to change and evolve with technology, from typewriters to calculators and the internet. What has remained true with all these changes is the increasing value of the what and the why, and the decreasing value of the how. This trend continues with AI.

At the end of the day, LLMs are fundamentally incredibly powerful predictors of the next word given a set of prior words. What the various model companies (e.g., OpenAI) have been able to do with that prediction is amazing. However, I fundamentally believe that AI is unable to perform critical thinking. The human's role using the AI is to evaluate and set the bar for what constitutes good, and identify the problems to use AI to solve.

Critical Evaluation is about defining what good is and developing skill sets for all on using AI as a tool.