P R O J E C T S
Psychopathy and Bribery
Stage: Closed
Description/Highlights: 1) Expecting social sanctions does not reduce the frequency of bribe offers; 2) Variations in acceptance and detection rates overshadow the effect of expecting social sanctions; 3) Psychopathy level is associated with the frequency and probability of offering a bribe; 4) Psychopathic individuals are less self-regarding if the choice directly negatively affects another individual.
Predicting Risk-Taking
Stage: Closed
Description: Consider a lottery with two stages. The first stage has a 90% probability of winning and advance to the second stage, which has a 22% probability of winning. The prize for completing the two stages is five gold bars and zero for failing any stage. There’s also a safe alternative to the lottery that earns only one gold bar. Which of the two options will a decision-maker (DM) choose? This project evaluates the ability of utility-based models and machine learning models to predict out-of-sample observations. The project finds that subjective probability weighting models cannot sufficiently predict behavior. In comparison, a tree-based machine learning model that integrates an eye fixation metric predicts significantly better but with limitations.
AI-Induced Corruption
Stage: Data Collection
Description: Leveraging the predictive power of the best model of the eye-tracking project, this project tests the effect of human-AI interaction on the choice to bribe. Those who are predicted to bribe will receive advice from an AI agent that is meant to inhibit the behavior. Those who are predicted to not bribe receive advice that is meant to induce bribery.
LLM as Financial Analyst
Stage: Planning
Description: This project uses a group of LLM-agents to take the role of financial analysts. The agents will predict the financial performance of firms based on historical financial statements. A new set of financial statements will be presented to the agents after predictions for the year are submitted, and after learning the predictions of the human counterparts. This allows for reinforcement learning, which mimics the process human analysts learn from their individual performance relative to the crowd and the market.
Exploring Happiness Dynamics
Stage: Planning
Description: Focusing on SDG 3 (Health and Well-being), this project explores the dynamics of happiness and well-being in developing countries like the Philippines by utilizing a daily survey of activities and the emotions related to each activity. We aim to test how external and personal factors like socio-economic status, interpersonal interactions, state of public utilities, etc. influence variations in emotions and well-being on average.