Uncritical, Inconsistent, or Critical? University Students’ Beliefs About AI-Generated Knowledge

• In Event: AERA Roundtable Session Thursday 2:15 pm - Gold 4
  In Roundtable Session: Critical Perspectives on AI: Ethics, Equity, and Educational Futures

Abstract

Generative AI tools increasingly serve as information sources for university students, yet little is known about how students judge AI credibility. Surveying a representative sample of 1,289 university students, we measured general and task-specific epistemic beliefs about generative AI and identified belief profiles via two-step cluster analysis. Three qualitatively distinct groups emerged: Uncritical (high trust in generative AI, low verification; 40%), Inconsistent (skeptical in general, yet trusting specific AI-generated output; 38%), and Critical (low trust, high need for verification; 22%). Profiles differed in AI use frequency and purpose, with Uncritical students relying on ChatGPT the most. Findings highlight the need for early AI-literacy instruction that fosters verification skills and supports responsible integration of generative AI in classrooms and policy.

Authors

• Marek Urban, Czech Academy of Sciences

• Helena Beranová, Czech Academy of Sciences

• Cyril Brom, Charles University

• Jiří Lukavský, Institute of Psychology, Czech Academy of Sciences

• Filip Děchtěrenko, Charles University

• Veronika Hein, Czech Academy of Sciences

• Filip Svacha, Charles University

• Kamila Urban, Slovak Academy of Sciences