Fostering the adoption of AI chatbot advisory for financial recommendations: The moderation of algorithmic interpretability

Abstract

AI chatbots are increasingly credible in using algorithms to propose personalized investment recommendations, financial plans and strategies. In this context, scholars are also interested in ethical, biased issues and deviations in the recommendations. This study proposes an empirical model to test the relationship between recommendation features (unbiased, quality, and quantity), users’ perceptions of algorithmic ethical issues (transparency, fairness, and accountability), algorithmic credibility, and users’ behavioral intentions when using AI chatbots in the financial sector. In particular, we also test whether algorithmic interpretability can moderate the relationship between ethical issues, credibility, and users’ intentions to use it. Data were collected from 419 respondents through an online survey questionnaire and analyzed using a PLS-SEM model. Research shows that providing unbiased recommendations improves users’ perceptions of fairness, transparency, and accountability. Additionally, the quantitative of recommendations provides a new perspective that more recommendations can benefit users’ perceptions. The study also extends the discussion by directly linking recommendation quality to perceived fairness. Furthermore, we found that users’ perceptions of algorithmic fairness impact on the credibility in AI chatbots, which positively influenced the behavioral intention to use chatbots for financial recommendations. However, our study shows that algorithmic interpretability does not influence any of the relationships between the above variables. Our research offers an enabling, fresh perspective on applying AI chatbots in finance. This not only increases user trust but also accelerates adoption rates, contributing to a fairer and more inclusive financial industry.