Calibrated Explanation-Detail Transparency in AI Recommendation Explanations: Trust, Privacy Concern, and Purchase Intention in E-Commerce

Abstract

Artificial intelligence (AI)-based recommender systems increasingly provide explanations for product recommendations in e-commerce. Although such explanations can reduce algorithmic opacity and strengthen consumer trust, they can also make data collection, profiling, and personalization practices more salient. This study develops and tests a calibrated explanation-detail transparency model in which recommendation explanations serve as both competence signals and privacy-risk signals. We conducted a one-factor, three-level, between-subjects scenario experiment with 472 valid responses retained from 623 submissions. Participants were randomly assigned to low-, medium-, or high-detail explanation conditions for the same product scenario. The data were analyzed using regression, PROCESS serial mediation analysis, covariance-based structural equation modeling, alternative-model comparisons, robustness checks, and statistical power sensitivity analysis. Higher-detail explanation conditions were associated with greater perceived algorithm understandability and platform trust, providing evidence consistent with a positive trust-benefit mechanism. They were also associated with greater perceived data intrusiveness and information privacy concern, providing evidence consistent with a negative privacy-cost mechanism. Purchase intention exhibited an asymmetric concave pattern across the three discrete conditions and was highest in the medium-detail condition. These findings support calibrated, layered, and privacy-sensitive explanation design in AI recommendation interfaces.

Keywords: AI recommendation, explanation-detail transparency, recommender systems, platform trust, information privacy concern, purchase intention, electronic commerce