Technology AI, Science and New Things
A study by Reading University has found that AI language models such as ChatGPT got better results in last year’s psychology exams than 80 per cent of its real undergraduates. The bot’s use was also rarely detected, with just one in 20 of the AI submissions being flagged as suspicious. Peter Scarfe, the university psychologist who conducted the experiment, described the findings as a wake-up call. The study raises concerns about cheating and remote exams, which were common during Covid but are still used by most British universities, including Oxford. Scarfe said AI did especially well in the first and second-year exams, but struggled more in finals. “There has been quite a lot of talk about the use of so-called AI detectors, which are also another form of AI,” he added, “but [the scope here] is limited”. The 33 fake students who submitted AI answers passed with honours, ending up, on average, half a degree classification higher.
