Algorithm aversion revisited: The role of AI literacy and attitudes towards AI in shaping perceptions of AI ‐generated texts

Abstract Scientific publications on AI education frequently express concerns that students at all educational levels, lacking sufficient AI literacy, may become passive learners due to the use of generative language models and blindly trust AI outputs. Concurrently, recent research has increasingly identified an ‘algorithm aversion’ tendency, leading individuals to regard information generated by AI with scepticism. Both uncritical trust and unfounded aversion can affect the efficient use of AI‐generated educational content. In an online experiment, participants assessed the credibility, usefulness and comprehensibility of AI‐generated text summaries labelled as AI‐, human‐ or hybrid‐generated. Validated instruments were employed to assess AI literacy and attitudes towards AI. Participants rated the credibility and usefulness of AI‐generated text excerpts that were explicitly labelled as AI‐generated significantly lower than AI‐generated texts that were labelled as human‐written or as the result of human–AI collaboration. However, this effect was relatively small, as all texts received highly positive evaluations. Furthermore, individual attitudes towards AI appeared to influence the assessment of AI‐generated texts. Incorporating potential moderator variables such as attitudes towards AI may help contextualize the sometimes contradictory findings on algorithm aversion and algorithm appreciation. Both pro‐ and anti‐AI biases could have substantial practical implications for the use of AI technologies in education.