Humans often behave in seemingly irrational ways. A common instance of such perplexing behavior is that we typically care about how and why people chose their actions, rather than caring only about the actions themselves. For example, when people agree to do us a favor, we prefer them to do so directly, rather than to first gather all the relevant information. Using game theory, we show that this preference may in fact be rational: The decision-making process often reveals hidden preferences of the decision maker, which can become relevant in a future interaction. This work elucidates the conditions that make caring about motivations beneficial and makes predictions regarding the real-world situations in which it is expected to occur.It is common sense that costs and benefits should be carefully weighed before deciding on a course of action. However, we often disapprove of people who do so, even when their actual decision benefits us. For example, we prefer people who directly agree to do us a favor over those who agree only after securing enough information to ensure that the favor will not be too costly. Why should we care about how people make their decisions, rather than just focus on the decisions themselves? Current models show that punishment of information gathering can be beneficial because it forces blind decisions, which under some circumstances enhances cooperation. Here we show that aversion to information gathering can be beneficial even in the absence of punishment, due to a different mechanism: preferential interactions with reliable partners. In a diverse population where different people have different—and unknown—preferences, those who seek additional information before agreeing to cooperate reveal that their preferences are close to the point where they would choose not to cooperate. Blind cooperators are therefore more likely to keep cooperating even if conditions change, and aversion to information gathering helps to interact preferentially with them. Conversely, blind defectors are more likely to keep defecting in the future, leading to a preference for informed defectors over blind ones. Both mechanisms—punishment to force blind decisions and preferential interactions—give qualitatively different predictions, which may enable experimental tests to disentangle them in real-world situations.