Cognitive models are a fundamental tool in computational neuroscience, embodying in software precise hypotheses about the algorithms by which the brain gives rise to behavior. The development of such models is often largely a hypothesis-first process, drawing on inspiration from the literature and the creativity of the individual researcher to construct a model, and afterwards testing the model against experimental data. Here, we adopt a complementary data-first approach, in which richly characterizing and summarizing the patterns present in a dataset reveals an appropriate cognitive model, without recourse to an a priori hypothesis. We apply this approach to a large behavioral dataset from rats performing a dynamic reward learning task. The model revealed suggests that behavior on this task can be understood as a mixture of three components with different timescales: a quick-learning reward-seeking component, a slower-learning perseverative component, and a very slow "gambler's fallacy" component.

Recent years have seen a surge of research into the neuroscience of planning. Much of this work has taken advantage of a two-step sequential decision task developed by Daw et al. (2011), which gives the ability to diagnose whether or not subjects' behavior is the result of planning. Here, we present simulations which suggest that the techniques most commonly used to analyze data from this task may be confounded in important ways. We introduce a new analysis technique, which suffers from fewer of these issues. This technique also presents a richer view of behavior, making it useful for characterizing patterns in behavior in a theory-neutral manner. This allows it to provide an important check on the assumptions of more theory-driven analysis such as agent-based model-fitting.