Reports of AI-generated work by students are piling up. The media are debating how educational institutions are supposed to tell which texts were actually written by people.
The problem
As language models and prompting methods improve, clear detection becomes increasingly difficult. AI detectors are not reliable themselves – and frequently misidentify texts by non-native speakers as AI-generated.
Possible approaches
Some institutions are taking new routes: oral examinations, project work instead of written assignments, or explicitly declaring AI use as a permitted source.
Till Krause of Landshut University allows students to use AI actively as a source – as long as this is clearly declared.
Our assessment
Academic assessment procedures will have to change. AI – like other technologies before it – has already arrived in students' everyday lives. What is needed now are ways of testing human knowledge differently.
