Model-based control predicts financial behaviour change: a randomised behavioural trial of habit substitution and goal-setting assessed using computational phenotyping

The dual-process framework of habit and goal-directed control provides a fundamental lens for understanding financial decision-making, yet the relative effectiveness of interventions targeting these distinct systems has not been directly tested using computational markers. We compared two theory-based strategies in a 12-week behavioural randomised controlled trial with students at two universities in the United Kingdom and Australia ( n = 224 completers from 397 enrolled). The Goal-Setting intervention leveraged model-based control through expense tracking, explicit savings targets, and progress feedback. The Habit-Based intervention targeted model-free processes through behavioural substitution by prompting cheaper alternatives for routine purchases. Using objective bank transaction data, the Goal-Setting intervention significantly reduced monetary consumption and improved account balances compared to control. The Habit-Based intervention showed no significant effects. Computational phenotyping revealed that individual differences in model-based control, measured through a two-step sequential decision task, predicted intervention success: participants with stronger goal-directed control (higher w parameter) showed greater spending reductions in both interventions. These findings demonstrate that while goal-directed scaffolding produces robust financial behaviour change, individual differences in model-based control predicted intervention success. The Goal-Setting effect on consumption remained significant in the smaller subsample that provided both spending and balance data, though it was estimated there with reduced precision. Our results indicate that habit-focused interventions may require stronger environmental restructuring than simple substitution prompts. Pre-intervention computational phenotyping could enable personalised financial interventions.