Using AI and Behavioral Finance to Cope with Limited Attention and Reduce Overdraft Fees

Abstract

In a field experiment using Mint, a personal financial management application operating in the United States and Canada, we investigate mechanisms to reduce overdraft fees. A sample of users identified via an AI algorithm developed by Mint as having a propensity for overdraft were sent alert notices to test the efficacy of the different framings in reducing the number of overdraft fees. We employ parametric identifications, as well as time-to-event semi-parametric analysis to learn that sending a reminder proved effective in and of itself, and the impact was significantly enhanced by simplifying the message. A negative framing of the simplified version elicited greater engagement and had stronger impact than a positive framing. Significant effects were obtained predominantly among the population with medium to high annual incomes. We relate our findings to the literature on limited attention and the ostrich phenomenon. Our work also contributes to the literatures on fintech, artificial intelligence, and human interaction.