Thinking Fast, Trading Slow: Confidence and Risk in Digital Investment Decisions

Stock market investors increasingly make consequential decisions through digital platforms that control what market information is available, how much it costs, and when performance feedback is delivered. How these platform features independently shape investment performance and belief accuracy remains poorly understood. We designed Market Mindset, a purpose-built web-based investment simulation, and tested these features in a 3×2 between-subjects experiment (N = 223) crossing information pricing (fixed cost, variable cost, no information) with feedback timing (per-round versus end-of-task). Feedback timing, but not information pricing, was significantly associated with confidence calibration in a secondary ANOVA on Brier scores (p = .021): participants receiving per-round outcome feedback were better calibrated than those in the end-of-task feedback condition. A post-hoc exploratory analysis further found that per-round feedback halved the odds of purchasing paid market information (OR = 0.50, p = .011), a pattern grounded in rational inattention theory that replicates across analytic specifications. Overconfidence was large and pervasive: 85\% of participants overstated their predictive accuracy (d = 1.01). Participants with information access did not outperform those without it; after information costs, they earned less on net. These findings indicate that feedback design, rather than information pricing structure, more reliably shapes belief accuracy in repeated investment tasks.