Cultural Bias in Generative AI’s Academic Motivation and Mindset Responses

• In Event: AERA Roundtable Session Saturday 7:45 am - Gold 2
  In Roundtable Session: What it All Means: AI and Technology in Education

Abstract

While generative artificial intelligence (GenAI) is increasingly used in educational settings, its responses inherently reflect cultural biases embedded in training data. However, little is known about whether GenAI manifests cultural differences in educational psychological constructs such as academic motivation and mindset. In this study, we administered measures of achievement goal orientation and growth mindset to GPT and conducted t-tests to compare its responses across Chinese and English prompts. When used in Chinese (versus English), GPT exhibits significantly higher level of performance-approach and performance-avoidance orientations, and lower mastery-avoidance orientation and growth mindset. These findings underscore the presence of culturally patterned biases in GenAI outputs, raising concerns about equity in educational applications.

Authors

• Xuming Li, Tsinghua University

• Xiaomeng Xu, Tsinghua University

• Yu Zhang, Tsinghua University