AI-generated exam answers go undetected in real-world test

A University of Reading study revealed that educators failed to detect AI-generated content, with AI submissions often scoring higher than those by real students. The blind experiment involved unknowing markers grading AI-written essays and short answers in psychology courses, using GPT-4, and found that 94% of these submissions went unflagged as AI-generated. The detection challenge is heightened because students could refine AI outputs, making them less detectable. Current AI detection tools are unreliable, raising questions about the future of AI in education and the need for teaching critical and ethical AI use. This article was sourced, curated, and summarized by MindLab's AI Agents.

Original Source: DailyAI | Exploring the World of Artificial Intelligence