The standard way of thinking about emotions in cognitive science starts with their function. The function of the fear program, for instance, is to help the individual evade imminent dangers. This functionalist proposal illuminates the character of the fear program, e.g., the kinds of things that elicit fear, and the kinds of responses that fear produces. The functionalist approach has been extremely productive, but it faces a puzzle with the emotion of guilt, for it’s unclear what function the guilt program serves for the individual. As Deem & Ramsey put it: “It seems that it is good for you that others are guilt-prone…, it is less clear that being guilt-prone is good for the individuals themselves” (2016, 571). Extant functionalist attempts to solve this puzzle (e.g., Frank 1988) have important shortcomings. To resolve the puzzle, we argue that the functional approach has been overly restrictive. Some cognitive systems need to be understood in terms of the functions those systems serve, not for the individual himself, but for others. That is, some cognitive systems have functions that are external to the individual. Just as the function of an artifact needs to be understood in terms of the interests of the artisan, so too the function of some cognitive systems needs to be understood in terms of the interests of those (e.g., parents, partners, or teachers) who crafted or shaped the cognitive system. This provides an alternative way of thinking about the function of the guilt program. On the external approach, we need to consider the function of guilt from the perspective of those who installed or edited the guilt program in the individual. From that perspective, it’s plausible that a primary function of the guilt program is precisely to protect the individual(s) who stand to be harmed by the agent’s action.