Critical Trust in Generative AI

Abstract

In a 2025 international study involving 11,706 undergraduate students, 29% overall (67% in the U.S.) reported regular use of generative artificial intelligence (GenAI) programs. This could be problematic, since GenAI outputs often contain randomly parroted falsehoods, and students could overuse the technology. This manuscript proposes a new framework, critical trust, to help university students balance trust and mistrust in the validity of information in GenAI outputs. A critical trust framework could also help students make ethical decisions about how and when to use GenAI applications in their academic work. University students can employ critical trust in critical friend groups, which could be especially beneficial for students from historically marginalized communities without college-educated mentors, helping them avoid accusations of misuse of GenAI and the loss of credibility that can result from plagiarism. The critical trust framework could also provide education researchers with a pathway to develop and test a scale that measures critical trust, ultimately determining the optimal levels to minimize harms from GenAI hallucinations and inappropriate overuse.

Ähnliche Arbeiten