Abstract
INTRODUCTION: Grounded in Behrens et al.'s (2008) advice-taking paradigm, this study investigates how advice phrasing (positive vs. negative) and task framing (gain vs. loss) influence the extent to which individuals integrate advice during decision-making. Rather than focusing on isolated choice outcomes, we examined the cognitive processes underlying advice use through a reinforcement learning (RL) framework. METHODS: Across two experiments (N = 38 and N = 74), participants completed probabilistic decision-making tasks while receiving trial-by-trial advice. Computational modeling was used to estimate the latent advice reference weight (ω), reflecting reliance on advice throughout the learning process, as well as the advice-specific learning rate (α (a) ). Behavioral measures of advice-taking (advice-choice consistency) were analyzed alongside modeling-derived parameters. RESULTS: Both behavioral indices and parameter estimates showed that participants relied more on positively phrased advice than negatively phrased advice. Moreover, advice phrasing interacted with task framing: positively phrased advice exerted a stronger influence under the gain frame, whereas negatively phrased advice was more influential under the loss frame. This interaction was robustly captured by the modeled advice-weight parameter (ω), although not consistently evident in behavioral choice patterns. Modeling results further showed that the advice-specific learning rate (α (a) ) was significantly higher for positively phrased advice, suggesting greater updating from such information. DISCUSSION: These findings provide a mechanistic understanding of how social (advice phrasing) and contextual (task framing) features jointly shape advice integration and inform more effective communication strategies in decision-making contexts.