Analytic Structuring, Psychological Empowerment, and Self‐Reported Trust Regulation in AI‐Framed Decision Support: Evidence From a Scenario‐Based Experiment

Abstract

Artificial intelligence (AI) increasingly supports human decision making by generating recommendations and structured evaluations. Yet how well an AI‐supported decision aid works depends on more than algorithmic accuracy; the surrounding interface shapes how individuals read and act on the advice it delivers. This study asks how analytic structuring—presenting a decision problem through an explicit multi‐criteria structure—relates to task‐level psychological empowerment, self‐reported trust regulation, and decision quality when people evaluate advice from an AI‐framed decision aid. Drawing on behavioral decision research on advice taking and on trust in automation, we propose the analytic structuring–empowerment–trust (AET) framework. A scenario‐based experiment used a 2 × 2 between‐subjects design that manipulated analytic structuring (high AHP‐based structuring vs. low structuring) and AI explanation transparency (high vs. low). Of 271 management students recruited at a Taiwanese university, 21 were excluded according to predefined screening rules, yielding a final sample of 250. Participants interacted with a scripted decision‐support interface that presented a fixed, pre‐programmed AI recommendation for an entrepreneurial decision task and responded to validated survey measures. The results fit the proposed framework: Analytic structuring was positively associated with psychological empowerment ( β  = 0.37, p  < 0.001), which in turn predicted perceived decision quality ( β  = 0.46, p  < 0.001). A bootstrapped analysis indicated a significant indirect effect of analytic structuring on decision quality through psychological empowerment (indirect effect = 0.17, 95% CI [0.11, 0.24]). Self‐reported trust regulation positively moderated the relationship between empowerment and decision quality ( β  = 0.19, p  < 0.01), and analytic structuring was itself positively associated with trust regulation ( β  = 0.33, p  < 0.001). Because the decision aid was scripted and its reliability never varied, these findings concern how the structured presentation of algorithmic advice shapes users' experienced agency and their self‐reported regulation of reliance, not behaviorally observed trust calibration. The contribution to behavioral decision research is the demonstration that how advice from an AI‐framed aid is structured, and not merely whether its reasoning is made visible, relates to how individuals experience and evaluate the decision process, with design implications for interfaces that support user agency and reflective engagement with algorithmic advice.