Research on factors affecting trust formation of generative AI agents in human–AI interaction contexts

Engineering notes

Notably, perceived reliability significantly mediated the associations between these two predictors and human-agent trust, whereas the indirect association involving perceived sense of agency approached but did not reach the conventional significance threshold and was therefore interpreted only as suggestive evidence.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Introduction As generative artificial intelligence (GenAI) becomes deeply embedded in cognitive workflows, its inherent probabilistic generation mechanisms and the possibility of machine hallucinations may complicate the stability of semi-technical users’ trust. This study examined how semi-technical users’ interaction trust toward generative AI agents is associated with perceived explainability, perceived intention alignment, and perceived sense of agency, while also considering the moderating roles of perceived task complexity and domain self-efficacy. Methods Survey data from 312 semi-technical users with experience in AI-assisted development were analyzed using structural equation modeling. Results The results indicated that perceived explainability and perceived intention alignment were strongly and positively associated with perceived reliability, whereas perceived sense of agency showed only a small positive association with perceived reliability. Perceived reliability, in turn, was positively associated with human-agent trust. Notably, perceived reliability significantly mediated the associations between these two predictors and human-agent trust, whereas the indirect association involving perceived sense of agency approached but did not reach the conventional significance threshold and was therefore interpreted only as suggestive evidence. Furthermore, perceived task complexity positively conditioned the relationship between perceived explainability and perceived reliability. Domain self-efficacy also showed a significant positive moderation pattern in the relationship between perceived intention alignment and perceived reliability, which was opposite to the hypothesized negative direction. Discussion These findings suggest that transparent and intention-aligned interaction designs may be especially important for supporting reliability-based trust formation in generative AI-assisted development, while user agency may require additional mechanisms beyond perceived reliability to account for its role in trust formation.

Links and sources

• PDF from original source

Comments