Expertise, Numeracy, and AI Trust in Financial Decision Making: When Calculation Strategies Meet Deliberation Time

Financial decision-making is increasingly influenced by external advice, raising questions about how individuals calibrate trust in human experts versus artificial intelligence. To make decisions consistent with normative standards like Expected Value (EV) in monetary lotteries, individuals must actively perform calculations rather than relying on intuitive heuristics. However, cognitive limitations such as low objective numeracy or cognitive depletion often hinder their ability to execute these analytical strategies.This study tested whether brief instructional interventions, framed by different authoritative sources, could enhance financial decision-making in monetary lotteries. Participants were randomly assigned to one of four conditions: Control, General Introduction, Expert, or AI. In the Control condition, participants made choices across 40 monetary lotteries without strategic guidance. In the other conditions, participants received identical instructions on computing EV prior to the task. These instructions were presented either neutrally (General Introduction), framed as advice from a human economist (Expert), or framed as a strategy from a personalized algorithm (AI).Results indicated that the general interventions did not universally increase EV-consistent choices compared to the Control group. However, important nuanced interactions emerged. Deliberation time mediated the relationship between the Expert condition and EV-consistent choices, as participants motivated by human authority slowed down to evaluate options. Conversely, high perceived objectivity in the AI condition actively suppressed task performance, suggesting an automation bias where unwarranted trust replaced analytical verification. Furthermore, higher objective numeracy positively correlated with normative choices, but its benefits were fully realized only by participants who maintained cognitive endurance throughout the task.In summary, the findings provide evidence that improving financial decision-making requires more than simple informational nudges or blind reliance on artificial expertise. While the overarching interventions had limited universal effects, optimal outcomes depend on a delicate synthesis of numerical competence, cognitive stamina, and calibrated trust. Materials and data for this study are available at: https://osf.io/hd94y/