People form and use inferences based purely on others’ simple affective responses to outcomes

Have you ever seen someone look disappointed after an outcome—whether a gift, a grade, or a rejection? Rich emotional reactions, such as disappointment, readily inform social inference (e.g., that the other person expected more). Yet real-world emotional reactions are not always rich, as they can take the form of simple indications of feeling good (e.g., feeling 8/10) or bad. Are such simplified affective reactions sufficient for social inference? Research on experienced affect and reinforcement learning suggests simple reports of feeling good or bad (i.e., valence) contain rich information: (i) people feel better when outcomes exceed expectations and worse when they fall short; and (ii) how people feel after an outcome predicts their subsequent choices. This raises the possibility that observed simple affective valence is already sufficient for social inference. Across four preregistered experiments (N = 783), participants observed two hypothetical demonstrators react to lottery outcomes via continuous valence reports. From these affective cues, observers correctly inferred which demonstrator expected more, while tracking trial-by-trial updating over time. Inferences generalised beyond the lottery context: demonstrators with higher expectations were rated as more dominant but less fair and likable, based solely on their reported feelings. Observers also adjusted their behaviour accordingly. They shared less with high-expectation demonstrators when generosity was voluntary, but offered more to them when rejection was possible. Together, our results establish simple affective valence as a sufficient signal for systematic social inference.