Identifying conceptual dimensions of trust in artificial intelligence from qualitative content analysis of open-ended responses
TL;DR
The findings indicate that lay conceptualizations of AI trust are predominantly anchored in performance expectations, but with significant and noteworthy secondary emphasis on moral attributes related to in safety and integrity.
Abstract
Trust in artificial intelligence (AI) has received growing attention in academic research, ethical debates, and societal discourse. However, limited research has examined how laypeople conceptualize trust in AI assistants. This study aims to identify recurring conceptual dimensions that underlie lay definitions of trust in AI. Open-ended survey responses were collected from English-speaking participants (N = 204) recruited via two online platforms. Responses were analyzed using inductive qualitative content analysis within a postpositivist framework. Three core themes emerged. The most prevalent dimension mentioned was performance (85%), with participants defining trust primarily in terms of accuracy, source quality, and competence. The second dimension emphasized safety (38%), reflecting participants’ concerns about harm prevention, protection and responsible system behavior. The third dimension involved perceived moral integrity of AI (26%), including references to unbiasedness, honesty, and fairness. The findings indicate that lay conceptualizations of AI trust are predominantly anchored in performance expectations, but with significant and noteworthy secondary emphasis on moral attributes related to in safety and integrity. These results offer an empirically grounded framework for understanding public trust in AI and may inform future measurement and design initiatives.