The Impact of Phrasing on Advice-Taking Under Gain and Loss Frames in a Reinforcement Learning Paradigm

Xuanhan Chang^*

• Department of Neuroimaging, School of Neuroscience, Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, United Kingdom

Grounded in Behrens et al.'s (2008) advice-taking paradigm, this study employed a reward-based reinforcement learning (RL) framework to examine how advice phrasing (positive vs. negative) and task framing (gain vs. loss) influence the extent to which individuals integrate advice into their decisions. Rather than analyzing isolated decision outcomes, we used computational modeling to estimate the latent reference weight (𝜔), which reflects the extent to which individuals relied on advice throughout the learning process. Across two experiments (N = 38 and N = 74), both behavioral and modeling results (specifically reference weight 𝜔) revealed that: (1) participants were more likely to follow positively phrased advice than negatively phrased advice; and (2) advice phrasing and task framing interacted such that positively phrased advice exerted a stronger influence in the gain frame, while negatively phrased advice was more influential in the loss frame. This pattern was robust in the modeled advice weight parameter (𝜔), but not supported by behavioral data. Computational modeling further showed that the advice-specific learning rate (αₐ) was significantly higher for positively phrased advice, suggesting greater updating from such information. These findings offer a mechanistic understanding of how social and contextual features shape advice integration and inform more effective communication strategies in decision-making settings.