Can AI Understand the Value of my Poetry Assignment? Critical Role of Perceived AI Task Fit and Outcome Favorability in User Trust

초록

Receiving a disappointing grade from an AI evaluator often leads individuals to question the system's competence. This external attribution is common and underscores why many remain uneasy about AI-based evaluations. Yet, what specifically fuels this distrust? This study investigates how receiving a negative evaluation affects people's trust in AI compared to human evaluators, particularly focusing on the perceived suitability of the evaluation task for AI. In an experiment with 448 young adults, participants were significantly more likely to blame AI evaluators for unfavorable outcomes, especially when tasks were subjective (e.g., poetry assessments) rather than objective (e.g., biology lab reports). Such blame considerably undermines trust in AI systems. This research highlights the importance of aligning AI's capabilities with appropriate evaluation tasks as a critical pathway to enhancing trust in AI-based evaluation systems.

키워드