THE IMPACT OF AI INNOVATION ON RETAIL INVESTORS’ DECISION-MAKING: THE MEDIATING ROLE OF SELF-ATTRIBUTION AND HINDSIGHT BIAS

Abstract

Background: The integration of artificial intelligence (AI) in the financial market has essentially recognized that how retail investors make decisions. AI-powered tools, such as robo-advisors and predictive analytics, can help improve the quality of investment decisions. However, cognitive biases—particularly self-attribution bias and hindsight bias—may still distort judgment, even in AI-assisted environments.

Objectives: The current paper examines how AI innovation affects investment decision making of retail investors by exploring the potential role of self-attribution and hindsight bias in mediating this association. Six hypotheses were tested to evaluate direct and indirect effects among AI innovation, behavioral biases, and decision.

Methodology: A cross-sectional quantitative design was employed under a positivist paradigm. Primary data were collected through structured questionnaires from 349 retail investors on the Pakistan Stock Exchange (PSX). Standardized, validated scales were used to measure AI innovation (Parasuraman, 2000), self-attribution bias, hindsight bias (Baker et al., 2019), and investment decisions (Gill et al., 2018). SPSS and Hayes’ PROCESS Macro (Model 4) were used to analyze reliability, correlations, regressions, and mediating effects between variables.

Results: The results indicated that AI innovation contributions raised investment decisions substantially and significantly affirming H1 (B = 0.124, p = 0.002). However, AI innovation did not significantly influence self-attribution bias (p = 0.771) or hindsight bias (p = 0.619), leading to the rejections of H2 and H3. Conversely, both hindsight bias (B = -0.209, p < 0.001) and self-attribution bias (B = -0.328, p < 0.001) proved to have significant negative impacts on investment decisions, supporting H4 and H5. In the case of H6a and H6b, mediation is conducted by way of Hayes PROCESS Macro (Model 4) showed that the indirect effects of AI innovation on investment decisions through self-attribution bias (effect = 0.0043) and hindsight bias (effect = 0.0060) were not significant, as their bootstrap 95% confidence intervals included zero ([H6a: -0.0210 to 0.0328]; [H6b: -0.0284 to 0.0379]). This indicates that neither bias significantly mediated the relationship, both H6a and H6b were rejected.

Conclusion: To sum-up AI innovation improves retail investors’ decisions directly, it does not reduce self-attribution or hindsight bias. These biases independently and negatively affect decision quality, underscoring the need for bias-mitigation strategies within AI systems.