The Problem of Trusting AI Is Not a Technology Problem but a Human Problem.
AI Over-reliance Arises Not from Ignorance but from Trust.

The explosive proliferation of recent generative AI is deeply penetrating human cognitive and psychological domains beyond technological progress. The study "An Integrated Exploration of Perceptions of AI-Generated Products and AI Over-reliance" by Kim Do-hee and Heo Chang-gu (2025) is noteworthy in that it empirically analyzed the relationship by positing AI as an active subject that interacts with humans and forms psychological attitudes. This study three-dimensionally illuminates through three studies the dual standards humans apply when viewing AI outputs, how certain personality traits lead to uncritical blind trust, and what psychological mechanisms AI dependence within the workplace combines with.

An Integrated Exploration of Perceptions of AI-Generated Products and AI Over-reliance

Kim Do-hee, Heo Chang-gu

Dual Evaluation by Generating Subject: Negative Bias Given by the Label 'AI'
Study 1 suggests that even outputs of the same quality are evaluated markedly differently depending on 'who made it.' An interesting finding is the reversal phenomenon of preferences appearing in the judgment domain (cover letter editing, medical diagnosis) and the information provision domain (travel schedule, stock investment). In the cover letter evaluation task, participants rated acceptability (F=5.66, p<.05), usefulness (F=6.67, p<.05), and trustworthiness (F=5.52, p<.05) all higher when the generating subject was human, regardless of output quality. Medical diagnosis also showed higher acceptability for human judgment.

On the other hand, in the stock investment recommendation domain where numbers and data are central, the opposite results appeared. AI outputs received higher scores than humans on accuracy (F=6.42, p<.05) and objectivity (F=20.25, p<.001). This shows that people trust AI as a 'sophisticated data processor' but still feel resistance to it as a 'subject of value judgment' where ethical responsibility or interpersonal empathy is involved. In particular, the fact that the interaction between quality level and generating subject was not significant suggests that the information itself that 'AI made it' acts as a kind of negative stigma causing evaluation bias, no matter how excellent the output.

Key Variables of AI Over-reliance: Individual Tendencies Rather Than Task Difficulty
Study 2 presents the insight that the phenomenon of 'AI Overreliance' — where users uncritically accept AI errors without recognizing them — is governed more by individual dispositional factors than situational factors. In an experiment using a maze-finding task, the correlation between easy task over-reliance and difficult task over-reliance was .70 (p<.001), appearing very high. This means people who over-rely on AI tend to consistently and blindly choose AI answers regardless of task difficulty.

Particularly unique is that attention was paid to the 'Honesty-Humility' factor of the HEXACO model beyond the Big Five personality model. Analysis results showed that the group that uncritically accepted AI errors in difficult task situations had significantly lower scores on 'Sincerity,' a sub-factor of Honesty-Humility, compared to the group that did not (t=2.14, p<.04), and their trust in AI was much higher (t=-2.69, p<.01). This empirically demonstrates that individuals with stronger desire to use the system for their own benefit and looser personal ethical verification mechanisms are at greater risk of accepting AI's wrong answers as-is as convenient tools.

Employed Workers' AI Dependence: Combination of Efficiency Perception and FOMO Anxiety
Study 3 analyzed multidimensional factors inducing AI Dependence targeting 200 current workers. AI dependence means a compulsive need state where one feels task performance is impossible without the system, going beyond simple utilization. Analysis results showed the factors most powerfully predicting AI dependence were AI usage time (r=.62, p<.001), AI utilization within work (r=.62, p<.001), and subjectively felt AI efficiency perception (r=.53, p<.001).

Notable is the association with emotional factors. Workplace FOMO (Fear of Missing Out, r=.40, p<.001) and depression (r=.39, p<.001) showed significant positive correlations with AI dependence. This implies that fear of falling behind in new technology trends or psychological contraction is operating as a mechanism increasing dependence on AI as a technological refuge. Also paradoxically, those with higher critical thinking tendencies (r=.18, p<.05) and higher need for cognition (r=.27, p<.001) showed higher AI dependence. This shows that individuals with analytical tendencies may deepen dependence in the process of more sensitively capturing AI efficiency and actively attempting to integrate it as their intellectual assets.

Need for Psychological Intervention Beyond Technology Management
This study clarified that AI over-reliance is not simply a problem arising from low technological understanding but a complex phenomenon intertwined with intimate personality traits and emotional dynamics within organizations. Interface design that clearly presents 'reasoning basis' for conclusions AI draws is only part of technical solutions. For actual AI management, not only education cultivating users' critical thinking but organizational-level psychological guidelines must be simultaneously accompanied so that individual psychological vulnerabilities like low sincerity or high FOMO anxiety do not lead to AI misuse.