LLMs are Biased Teachers: Evaluating LLM Bias in Personalized Education
This addresses bias concerns in AI-driven education, highlighting risks for diverse student populations, but is incremental as it focuses on evaluation rather than novel mitigation.
The study evaluated bias in large language models (LLMs) acting as personalized teachers, revealing significant biases in content generation across demographic groups like income and disability, with metrics showing potential harm to student learning through stereotypes.
With the increasing adoption of large language models (LLMs) in education, concerns about inherent biases in these models have gained prominence. We evaluate LLMs for bias in the personalized educational setting, specifically focusing on the models' roles as "teachers." We reveal significant biases in how models generate and select educational content tailored to different demographic groups, including race, ethnicity, sex, gender, disability status, income, and national origin. We introduce and apply two bias score metrics--Mean Absolute Bias (MAB) and Maximum Difference Bias (MDB)--to analyze 9 open and closed state-of-the-art LLMs. Our experiments, which utilize over 17,000 educational explanations across multiple difficulty levels and topics, uncover that models potentially harm student learning by both perpetuating harmful stereotypes and reversing them. We find that bias is similar for all frontier models, with the highest MAB along income levels while MDB is highest relative to both income and disability status. For both metrics, we find the lowest bias exists for sex/gender and race/ethnicity.