Presenter

Kelly Huang, MASc candidate in Systems Design Engineering

Abstract

Many older adults rely on family or friends to navigate digital banking, often sacrificing financial privacy or personal control. As autonomous AI enter consumer finance, they offer a way to manage money without relying on others, yet they also raise new questions about who holds decision-making authority. This thesis examines how an AI assistant’s autonomy and explanations shape older adults’ trust, perceived control, and understanding. In a mixed-methods, within-subjects pilot study, 33 adults aged 65–83 completed six simulated financial tasks with a scripted assistant that operated at three levels of autonomy: presenting options, recommending actions for approval, or acting with a 24-hour cancellation window. Each level was presented with and without explanations.

Varying levels of autonomy produced no detectable differences in trust, while perceived control remained high even when the assistant acted before seeking approval. Providing explanations significantly improved clarity and understanding without detectable gains in trust or perceived control. Qualitative findings helped explain these results, revealing that participants could value the utility of financial AI assistants without fully trusting them, and that they felt in control even when they had not authorized an action beforehand.

This thesis provides empirical evidence of how older adults experience autonomous financial AI and identifies a potential gap between perceived control and prior authorization. It also reveals challenges in interpreting trust measures after brief encounters, including reverse-worded items and participants’ readiness to judge. Together, these findings inform design considerations for consent timing, transparent system authority, user-defined delegation, and explanations that support informed financial decision-making.

Join us in EC4-2013!

This seminar counts towards the graduate student seminar attendance milestone!