Pavlovian bias instigates suboptimal choices in humans

Value-based decisions are often biased by Pavlovian cues in the environment, but it remains unclear how such biases affect optimal decision-making in humans. To address this, we used a novel variant of the Pavlovian-to-Instrumental Transfer (PIT) paradigm. Participants first learned to prefer a richer option (70 % reward probability) over a poorer one (30 %) in order to maximize rewards. A baiting rule was implemented, whereby each option's reward probability increased the longer it was not selected, rendering reward matching (i.e., aligning choice proportions with reward probabilities) a close approximation of the optimal strategy. Pavlovian cues predictive of either option were then introduced during decision-making. Results showed that Pavlovian bias disrupted optimal behavior, impairing reward maximization under both nominal extinction and rewarded conditions. This effect was replicated in an independent sample. In a third experiment, weakening the cue-outcome association during Pavlovian learning reduced the bias, suggesting a causal role of cue predictiveness. Together, these findings demonstrate that Pavlovian cues can produce maladaptive choice patterns, with implications for understanding and potentially mitigating behaviors linked to impulsivity and addiction.