DISC Research Seminar Series

During interactions with customers, service chatbots are increasingly equipped to provide choices to streamline the service delivery process. This research investigates when and why the implementation of choices enhances or impairs customers’ service experience. Based on the concept of fluency, we posit that the implementation of choices is beneficial only after a conversational breakdown due to a chatbot failure otherwise, the value of choice provision for facilitating fluency may not be salient enough. We further propose that choice provision is counterproductive when the choice set is incomprehensive, reducing (rather than enhancing) the fluency in the use of provided choices for a subsequent decision. We conducted several experiments to test these hypotheses. By illuminating when and why choice implementation may help or harm customers during a chatbot-initiated service interaction, we augment the current understanding of a chatbot’s role in customers’ service experience and provide insights for the deployment of choice-equipped chatbots in customer service.

We studied the effect of adopting an AI-enabled credit scoring model by a major bank on financial inclusion as measured by changes to the approval rate, default rate, and utilization level of a personal loan product for the underserved population. The bank served over 50 million customers and used a traditional rule-based model to evaluate the default risk of each loan application. It recently developed an AI model with higher prediction accuracy of default risk and used the AI model and the traditional model together in assessing loan applications for one of its personal loan products. Although the AI model could be more accurate in estimating default risk, little is known about its impact on financial inclusion. We investigated this question using a difference-in-differences approach by comparing changes in financial inclusion of the personal loan product adopting the AI model to that of a similar personal loan product without adopting the AI model. We found that the AI model enhanced financial inclusion for the underserved population by simultaneously increasing the approval rate and reducing the default rate. Further analysis attributed the enhancement in financial inclusion to the use of weak signals (i.e., data not conventionally used to evaluate creditworthiness) by the AI model and its sophisticated machine learning algorithms. Our finding is consistent with the statistical discrimination theory, as the use of weak signals and sophisticated machine learning algorithms improves prediction accuracy at the individual level, thus reducing the reliance on group characteristics that often lead to financial exclusion. We elaborated on the development process of the AI model to illustrate how and why the AI model can better evaluate the underserved population. We also found the impacts of the AI model heterogeneous across subgroups, and those with missing weak signals saw smaller improvements in the approval rate. A simulation-based analysis showed that simplified AI models could still increase the approval rate and reduce the default rate of the underserved population. We further discussed the compliance and generalizability issues about using AI and privacy-sensitive data in credit scoring. Our findings provided rich theoretical and practical implications for social justice by documenting how an AI model designed for improving prediction accuracy can enhance financial inclusion.