Universities rethink AI use as cheating checks prove unreliable

In short: Schools and universities are adjusting to wider AI use, as AI companies target education and some AI cheating detectors are found to be inaccurate.

What's going on

A new Financial Times special report describes how teachers and students are dealing with fast-growing use of AI tools in education. The report says AI labs like Anthropic and OpenAI are offering free or discounted products designed for classrooms.

At the same time, some universities are stepping back from AI “cheating detection” tools. These tools try to guess whether a student used AI to write an assignment, a bit like a plagiarism checker but for AI-written text. The report says schools have concerns about accuracy, meaning the tools can wrongly flag honest students or miss real misuse.

The report also highlights worries about “cognitive atrophy”, which is a drop in thinking skills from over-relying on help. It is similar to using a calculator for every small problem and then feeling rusty at basic math. One professor argues that universities should teach students “eval” skills, meaning the ability to test and judge AI outputs, rather than treating AI as always right.

What to watch

Expect more changes in how students are assessed. The report suggests some institutions are trying to move away from surveillance-style approaches and redesign assignments so students must show their thinking process. If AI tools keep spreading in classrooms, schools may put more focus on teaching students how to check AI answers and use them responsibly.

Source: Financial Times