Trust in AI versus human experts: a comparative study of user preferences between different educational AI applications

This study investigated the determinants of trust, AI aversion or appreciation, intention to use, and perceived usefulness of AI across eleven distinct use cases of educational AI applications. We surveyed 909 Finnish adults. Trust in AI varied significantly across educational applications. Notably, generative AI and chatbots were rated low in trust yet high in perceived usefulness and intention to use, indicating that high perceived usefulness can sustain intention to use despite lower trust, particularly among experienced users. Human experts were consistently trusted more than AI. Technical knowledge and a positive attitude toward technology emerged as particularly robust predictors of AI trust. Female respondents showed higher trust levels for both AI and human experts, differing from some previous findings in technology acceptance literature. We extend traditional technology acceptance models by revealing a ‘paradox of productive distrust’, providing a more nuanced understanding of the factors driving AI acceptance in education. Personality traits emerged as significant predictors of AI trust, with open-mindedness, emotional stability, and conscientiousness showing positive associations. Previous experience with AI applications emerged as a strong positive predictor of both trust and intended use, supporting the importance of actual interaction in building trust in AI systems.