Quick Buzz Feed
One AI tutor engaged middle schoolers. Another backfired with college students.
Research on the 'Goblins' math practice website used by 6,000 middle schoolers showed that consistent usage (at least two problems per week) correlated with significantly higher year-end math assessment scores, equivalent to about three additional months of learning. The tool successfully engaged a higher proportion of students, including English learners and those with disabilities, potentially due to its interactive, multilingual avatar and teacher control over assignments, thereby mitigating the common 'five-percent problem' in educational technology. However, the study notes that causality cannot be definitively proven, and higher engagement still only reached a small percentage for more intensive usage.
A separate experiment at the University of Maryland introduced an AI assistant into college courses for 1,100 students, while a control group of 1,200 students did not have access to it. Students in AI-assisted classes experienced a decrease in grades (an average of 0.11 points on a four-point scale) and reduced interaction with course materials, instructors, and peers. This negative impact was particularly pronounced for first-generation students. Only 15% of students used the assistant, primarily seeking answers rather than engaging with its tutoring features. Researchers hypothesize that the AI's presence inadvertently signaled permission for students to use external AI tools, potentially leading to less active learning and poorer academic outcomes.