Does AI help with homework but harm exam performance?

In this EdTechnical short, Owen is joined by guest co-host Alyssa Van Camp, research lead at EdTechnical, to discuss a recent paper on the learning penalty from generative AI. Drawing on survey data from around 27,000 secondary school students in China, the study looks at how using generative AI may affect homework productivity and learning. The headline finding is that students who use AI score higher on their homework, but they spend less time on it and their exam results drop. The findings fit with research on how effort shapes learning, though the study design has some limitations. Students self-reported their AI use, time spent on homework was measured indirectly through the school's learning platform, and the paper gives little insight into the quality of the exams used to assess learning. Surprisingly, around one in five students used AI without a drop in their exam scores. These students spent as much time on their homework as their peers, suggesting they used AI as a tool to help, rather than offload to. What does this mean for designing AI tools that support learning? And how might homework need to change? Links Research paper discussed in this episode: The Generative AI Learning Penalty: Evidence from Chinese Secondary Education Join us on social media: BOLD (@BOLD_insights), Libby Hills (@Libbylhhills) and Owen Henkel (@owen_henkel) Listen to all episodes of EdTechnical here: https://boldscience.org/podcasts/edtechnical/ Subscribe to BOLD’s newsletter: https://bold.expert/newsletter Stay up to date with all the latest research on child development and learning: https://bold.expert Credits: Sarah Myles for production support; Josie Hills for graphic design; Pia Schmoeckel for content production.