Game theoretic approaches to pragmatics characterize pragmatic inference as a product of interlocutors’ reasoning about their own and others’ possible linguistic choices. These choices are a function of their preferences over communicative outcomes. They are a product of what is communicated and how it is communicated, sometimes also factoring in beliefs about others’ preferences and beliefs. For instance, a speaker may prefer to bring a message across in a polite manner rather than in a more straightforward one, or succinctness may be preferred over long-windedness. At a more fundamental level, a speaker’s goal may be to convey his or her beliefs truthfully, or, the goal may be to deceive one’s audience. Hearers can have other, possibly opposed, preferences. Such subjective preference measures—captured by so-called utility functions—can thereby characterize a wide array of communicative scenarios, ranging from fully cooperative ones to ones with conflicts of interest. They also do away with the need to explicitly formulate conversational principles. Instead, pragmatic inference is directly rooted in interlocutors’ preferences and beliefs. Another cornerstone of game-theoretic approaches to pragmatics is that they often make the degree of mutual reasoning that interlocutors engage in explicit. While the simplest reasoners take only themselves as their reference point, more sophisticated ones iteratively reason about their partner’s choices and reasoning. A final key component common to these approaches is that they take a stance on interlocutors’ rationality: the degree to which they care about matters such as communicative success or manner. While some approaches assume full rationality, with interlocutors always acting according to what best fulfills their preferences, others weaken this assumption to allow for deviations. This makes them particularly suitable to predict and inform empirical data.