Digital Refuge or Digital Stress? How Perceived Usefulness of LLMs Reshape Academic Anxiety Through Identity and Self-Efficacy Pathways

• In Event: AERA Roundtable Session Wednesday 3:45 pm
  In Roundtable Session: Technology and Adolescent and Youth Development

Abstract

This study explores how high school students’ perceived usefulness (PU) of large language models (LLMs) affects academic anxiety. Grounded in TAM, SCT, and SIT, 1,670 students (aged 14–19) from six schools in Shanghai and Shandong were surveyed. Structural equation modeling revealed that PU of LLMs negatively predicted academic anxiety. Identity styles and academic self-efficacy (ASE) served as sequential mediators, with informational identity and higher ASE reducing anxiety. Gender moderated the PU-anxiety link, with males reporting higher PU and lower anxiety. These findings offer theoretical insights into affective responses to educational AI and inform gender-sensitive, psychologically adaptive strategies for AI-enhanced learning environments.

Authors

• Wencan Li, East China Normal University

• Xingchen Zhu, Liaoning Normal University