In social life, it is useful to identify social partners who are prosocial, supportive, and mindful of others. Past work shows how perceivers detect an agent’s social preferences (or their desire to benefit others) from directly helpful acts, such as generosity. However, perceivers might also detect an agent’s social preferences from indirect evidence, such as whether an agent’s personal choices are considerate of others’ known or possible preferences. Here we test a computational account of this capacity in which perceivers assume an agents’ personal choices aim to maximize their utilities, including the utility they assign to others’ outcomes, based on their knowledge. We tested our model across a series of pre-registered experiments which varied agents’ knowledge of others’ preferences (social knowledge) as they made personal choices. In Experiment 1a and 1b, participants inferred an agents’ social preferences based on personal choices which could reveal if the agent considered how their actions would indirectly affect others. In Experiment 2, perceivers were told the agent’s social preferences, and inferred their knowledge or ignorance about others’ preferences based on that agent’s personalchoices. Across both experiments, we find converging support for our computational account. We find that perceivers not only leverage information from an agent’s personal choices to infer their social preferences (Experiment 1a and 1b), but also deploy information about an agent’s social preferences to predict the agent’s knowledge from their made personal choices (Experiment 2). These findings illuminate how people can discern potential social partners from indirect evidence of their prosociality, thus deepening our understanding of partner detection, and social cognition more broadly.