Project 4 Outline: Leveraging AI-Driven Feedback to Mitigate Biases in Decision-Making

Supervisors

Dr Ozgur Akgun (School of Computer Science), Dr. Tugce Cuhadaroglu (Department of Economics)

Project description

This project aims to apply machine learning techniques to identify and reduce cognitive biases in decision-making.  Empirical evidence indicates that individuals frequently deviate from rational decision making by exhibiting phenomena such as violations of transitivity, overweighting small probabilities and underweighting large probabilities, loss aversion, overvaluing certain outcomes, inconsistent risky choices, vulnerability to framing effects, and differing risk preferences in gains versus losses, etc. The central question is whether these deviations stem from inadvertent errors in processing or are deliberate choices, and if AI-driven feedback can help identify and correct them.

The project is structured in two phases. First, theoretical models and predictions of economic decision making under uncertainty will be used to develop an algorithm that aims to estimate preference parameters, and likelihoods of potential biases from decision data. In the second phase, this algorithm will be used in an online experiment to test whether AI-driven feedback helps to mitigate those biases. The objectives are to determine if deviations from rational decision-making theories are due to mistakes in probability/information processing or deliberate choices, and in cases they are mistakes, to assess the efficacy of AI-driven feedback. This project aligns with the university strategy of promoting interdisciplinary research in marrying computational innovation with economic analysis, thereby enhancing our understanding of decision-making and offering practical insights into mitigating biases in uncertain environments.

Skills & Requirements

Applicants are expected to possess robust computer science skills, including experience in algorithm development and data analysis, alongside an interest in applying economic methods to behavioural research. Prior experience in machine learning is essential, as the project relies on the candidate’s ability to integrate advanced computational techniques with economic theory without the need for formal training in these areas.

Further details and application process

The project has been awarded St Andrews Business School Research Bursary of £1000 for research expenses. If you are planning to apply for this project, first contact the supervisors, Dr Ozgur Akgun ([email protected]) and Dr. Tugce Cuhadaroglu ([email protected]) to discuss eligibility and obtain approval. Once an agreement is reached, you can proceed to submit your formal application.