OECD report finds frequent AI use linked to lower student test scores

A recent OECD education report reveals that students who use AI chatbots infrequently or not at all tend to perform better academically than those who use them daily. Based on PISA data covering 91 countries and over 760,000 students, the study found that students who almost never use AI for writing assignments averaged 509 points in science tests, compared to 481 points for those using it daily. This gap represents approximately one and a half years of instructional difference.

Andreas Schleicher, Director for Education and Skills at the OECD, noted that learning requires active cognitive effort rather than passive reception. He suggested that AI should function as a “scaffolding” to enhance learning rather than a “crutch” that allows students to bypass necessary mental effort.

The relationship between AI and performance is nuanced. Students who use AI once or twice a week for learning purposes often outperform both non-users and daily users. Furthermore, students who are frequently required to critically evaluate AI-generated information in the classroom scored 13 points higher in science, indicating that the method of usage is a critical factor in academic outcomes.