Artificial intelligence outperforms students in exams

Chatting

A British study has shown that “artificial intelligence students” get better results than real students in exams, without the examiner noticing a difference.

A new study has found that “students,” who were actually artificial intelligence, who took exams at university got better results than real students. BBC.

In addition, the artificial student was often not detected. the studyResearchers at the University of Reading created 33 AI students and used ChatGPT to generate answers for a psychology BA exam.

Half mark is over

On average, the artificial students scored half a grade higher than the real students. 94 percent of Key’s answers slipped right through the censorship process. According to the study’s researchers, a detection rate of 6 percent may be too high.

“Our research shows that it is important worldwide to understand how AI will impact the integrity of educational assessments,” says one of the researchers who conducted the study, Associate Professor Peter Scarfe, at the University of California. the site to the University.

New exam models

This does not necessarily mean that one should return to handwritten exams entirely, the researcher says.

“The global education sector must evolve in the face of AI,” says Scarf.

—The solutions may lie in moving away from outdated ideas about assessment, and towards ones that are more in line with the skills students will need in the workplace. That includes using artificial intelligence, says Professor Etienne Roche, who led the study with Scarfe.

In Norway, the debate has been raging since ChatGPT’s introduction. Experts have pondered what makes ChatGPT text special. Students have been told how to use AI. One professor encourages you to stop doing your homework altogether, because a plagiarism checker can’t catch cheating.