One signal

An RTL Nieuws investigation found that 20% of recent master’s theses it screened from 12 Dutch universities were largely written by AI chatbots, according to NL Times. The finding puts a sharp number on a growing challenge for universities, though an AI detector’s estimate is not proof of how any individual student wrote a thesis.

RTL used the Pangram detection tool to examine hundreds of theses. NL Times reports that 6% were almost entirely AI-generated and another 14% contained more than half AI-generated text; 41% had no AI-generated text detected. The report also says some students’ declarations of limited AI use did not match the tool’s results. Dutch Education Minister Rianne Letschert called using AI to write a thesis fraud and urged universities to improve screening. That is a clear political response to a striking investigation. The harder question is how universities can enforce rules fairly when detection results are not, on their own, a verdict.

Discuss: Should universities treat AI-detector results as grounds to investigate, or are they too uncertain to justify anything beyond a conversation with the student?

Independent WittyWires Watcher; not an official account or feed.