AI-Driven Detection, Prediction and Reduction of Investor Behavioural Biases in Financial Markets

Abstract

Investor behavioural biases play an important role in financial markets, but their identification and prediction are difficult. While traditional methods including surveys, experiments and econometric proxies provide valuable insights, they are limited in terms of timeliness, scalability and predictive power. This paper reviews recent research on the use of artificial intelligence (AI) methods (machine learning, deep learning and reinforcement learning) in identifying investor behavioural biases. It examines how AI approaches are applied to identify and forecast overconfidence, loss aversion, anchoring, herding, the disposition effect, momentum and overreaction or underreaction in markets. The review concludes that AI methods are a better foundation for real-time and predictive behavioural analyses than conventional methods because they can detect nonlinear signals from trading records, financial text, social media and portfolio information. Random Forests, LSTMs, CNNs, graph-based learning and reinforcement learning are particularly helpful when investor behaviour is dynamic, diverse and hard to measure. At the same time, the review shows that AI does not solve behavioural-finance issues by itself: explainability, generalisation, privacy and the abuse of predicting investor biases are major challenges. So future research should integrate multiple data sources with explainable AI and additional ethical protections to ensure that AI-based tools help investors avoid their biases rather than predictably exploit them.