TRUST IN AI, BEHAVIOURAL NUDGING, AND INVESTOR OVERSIGHT IN RETAIL INVESTOR DECISION-MAKING: EVIDENCE FROM INDIA
TL;DR
The study's most significant finding is the non-support of H4: perceived investor oversight enhanced trust in AI but did not reduce bias susceptibility, indicating that oversight functions as a legitimacy signal rather than a debiasing mechanism.
Abstract
AI-powered investment platforms have fundamentally altered how retail investors in India access and act on financial information. Despite rapid platform adoption the psychological mechanisms through which AI recommendations, platform embedded behavioural nudges and perceived regulatory oversight shape investor decision-making remain poorly understood especially in emerging market contexts. This study addresses that gap by examining the direct and mediated pathways through which these three antecedents influence investment behaviour, operating through trust in AI and cognitive bias as dual mediating mechanisms. Drawing on the Technology Acceptance Model, Nudge Theory, Prospect Theory, and Trust-in-Automation literature, this study develops and tests an integrated structural model. Primary data were collected via a structured survey of 276 retail investors actively using digital investment platforms in India. Partial Least Squares Structural Equation Modelling (PLS-SEM) was employed to assess the measurement model and test eight hypotheses, including two mediation relationships, using bias-corrected bootstrapping with 5,000 subsamples. Five of six direct hypotheses were supported, with behavioural nudging emerging as the most powerful predictor of cognitive bias activation (β = 0.492, p < 0.001, f² = 0.324) — a medium effect that substantially exceeds the influence of AI recommendation quality. Both mediation hypotheses were confirmed: trust partially mediates the path from AI recommendation quality to investment behaviour (indirect β = 0.041, p = 0.036), while cognitive bias partially mediates the path from behavioural nudging to investment behaviour (indirect β = 0.070, p = 0.034). The study's most significant finding is the non-support of H4: perceived investor oversight enhanced trust in AI (β = 0.180, p = 0.002) but did not reduce bias susceptibility (β = 0.103, p = 0.096). This dissociation indicates that oversight functions as a legitimacy signal rather than a debiasing mechanism — investors feel more confident using AI-assisted platforms under visible regulation, but are not more rational in their decision-making. This is among the first studies to simultaneously operationalise AI recommendation quality, behavioural nudging, and perceived investor oversight as antecedents within a unified structural model, validated on Indian retail investor data. The oversight-bias dissociation also has direct implications for regulators: disclosure-oriented frameworks may build investor confidence without improving decision quality, suggesting the need for behavioural design standards alongside transparency mandates.