The Psychological Mechanisms and Social Implications of Over-Trust in AI Chat-Based on GPT-4 Persuasion Experiments

Abstract

As large language models have been increasingly applied to daily life in information seeking and decision-making, public trust in AI output has greatly improved. This brings about the worry that human critical thinking could become impaired. Using Salvi et al.'s GPT-4 conversational persuasion experiment as a start, this paper adopts a literature review method and relies on the theory of automation bias as well as cognitive laziness. This paper uses literature review as the way of research and utilizes the theories such as automation bias and cognitive laziness, etc., to explore the psychological mechanism behind the power of AI's ability to persuade others, and how it affects different people's minds. Findings indicate that the notable persuasiveness of AI stems primarily from two interrelated psychological mechanisms: cognitive laziness, which inclines individuals to forgo independent verification, and algorithm worship, which inclines individuals to delegate cognitive authority to AI systems. Both mechanisms manifest across adolescent and adult populations and may contribute to the erosion of critical thinking through prolonged AI use. The national, corporate, educational, individual levels response strategies are put forward in this paper, and the future research directions including cross-cultural comparison studies, longitudinal follow-up researches and individual-difference study plans are given out.