Reasoning goals and representational decisions in computational cognitive neuroscience: a case study from bounded evidence accumulation

Computational cognitive models are powerful tools for enhancing the quantitative and theoretical rigor of psychology and cognitive neuroscience. It is thus imperative that model users—those who develop models, use existing models, or integrate model-based findings into their own research—understand how these tools work, what limitations they have, and what factors need to be considered when engaging with them. To this end, we have developed φ-kit: a philosophical toolkit for informed, critical use of computational cognitive models in the behavioral and brain sciences. By compiling insights from philosophy of modeling and feminist philosophy of science, φ-kit explicates how the reasoning goals of a model builder fundamentally shape every step of the model-based research process. We demonstrate the utility of this philosophically-informed perspective by applying φ-kit to one of the most popular and successful family of models in cognitive science: accumulator models of two-alternative speeded decision making. The case study begins by investigating notions of optimality both as a general concept and as formalized in accumulator models. Then, we show how different commitments to model optimality (i.e., different reasoning goals) led to development of two “competing” standard forms of the diffusion model and offer a principled heuristic for deciding which form of the model to use. Finally, we demonstrate how insights from φ-kit, and the philosophical method more broadly, inspired the development of a novel theory about integration of expectations into decisions that synthesizes previously disparate findings into a common conceptual framework, thus concretely demonstrating how philosophy can advance neuroscientific practice.