Beyond Usefulness: The Role of Transparency and Explainable AI in Shaping Purchase Intention in Recommender Systems

Abstract

Algorithmic explanations may inform consumer evaluation, but understanding a rationale and trusting its source are distinct judgments. Drawing on Signaling Theory, this study develops a conditional framework in which explanations provide useful information when they are comprehensible, decision-relevant, and credible. A cross-sectional convenience survey of 208 e-commerce users was analysed using PLS-SEM to examine associations among perceived transparency, personalization, trust, usefulness, and purchase intention. Reported estimates include transparency–trust (β = 0.794; f2 = 1.71) and transparency–purchase intention (β = 0.350). The usefulness–purchase intention path was not statistically significant (β = 0.115, p = 0.113). The model accounted for 46.5% of purchase-intention variance. The unusually large transparency–trust association remains unresolved because complete measurement and common-method diagnostics are unavailable. Task complexity, perceived decision risk, and platform heterogeneity are specified as untested boundary conditions. The contribution is a set of falsifiable theoretical expectations accompanied by preliminary survey associations; the data do not validate the proposed mechanisms. Complete primary-output recovery or an independent benchmark study is required before the structural findings can support stronger substantive conclusions.