Abstract

In recent years, the widespread adoption of artificial intelligence (AI)-based tools in financial markets has fundamentally transformed investors' decision-making patterns. Retail investors, who are often confronted with cognitive limitations and behavioral biases such as overconfidence, anchoring, loss aversion, and herding behavior, are more affected by these developments than institutional investors. Examining the impact of AI on the financial decision-making efficiency of this group is highly significant, as their decisions influence the overall efficiency of the capital market.

The present study employed a quantitative approach with a survey method. The statistical population consisted of active individual investors in the Tehran Stock Exchange (TSE) and Iran Fara Bourse who had at least one year of trading experience and prior use of AI tools (such as investment advisory robots or smart applications). Convenience sampling was used, and from 538 distributed questionnaires, 500 valid cases were selected for final analysis. The data collection instrument was a standard questionnaire with 29 items based on a five-point Likert scale, measuring seven key constructs of the model (AI capabilities, information quality, transparency, trust in AI, reduction of behavioral biases, over-reliance, and decision-making efficiency). Data analysis was performed using structural equation modeling via the partial least squares (PLS-SEM) method in SmartPLS software version 4, with 5,000 bootstrap replications.

The findings revealed that AI technical capabilities (β=0.291) and provided information quality (β=0.380) have positive direct effects on decision-making efficiency. Reduction of behavioral biases showed the strongest positive path (β=0.427), while over-reliance had a significant negative effect (β=-0.175). AI transparency exhibited the strongest influence on trust (β=0.682), and trust had a weak positive direct effect (β=0.220) but a negative indirect effect (through over-reliance) on efficiency. The model's coefficient of determination for the decision-making efficiency construct was 0.615, confirming strong explanatory power.

The main innovation of this research lies in the simultaneous integration of psychological factors (trust, over-reliance, and biases) with AI technological features in a coherent empirical model tailored to the Iranian market. This model provides a foundation for future research and assists policymakers, regulatory bodies, and designers of smart tools in harnessing AI benefits while mitigating behavioral risks—for instance, by incorporating algorithmic transparency mechanisms and anti-over-reliance warnings in financial platforms.