Everyday decisions require leveraging different sources of information processed with complementary cognitive systems. For example, a decision about what to cook for dinner may rely on learning systems to supply meal preferences, memory systems to recall ingredients that are available, and perceptual systems to evaluate the quantity and quality of those ingredients. Each system is optimized to solve a particular type of problem, be it making accurate predictions based on prior experience, recalling decision-relevant information, or precisely representing the state of the sensory world. Despite having different goals, these systems rely on common computational principles to help achieve them. Effectively solving any of these problems requires pooling relevant sources of information to improve precision, but partitioning unrelated sources of information to avoid interference. Our research program focuses on these shared computational principles to better understand how their application in modular information-processing systems impacts decisions and complex behavior.