Survival requires proper association of reward outcomes to choices and actions that precede them. To learn successfully, however, the brain must constantly adjust how it incorporates each reward feedback because the relationship between choices and reward outcomes can unpredictably change over time. Moreover, choice options in the real world consist of many features or attributes and thus, additional information in the environment should be used to adjust what should be learned and ultimately, how learned and external information should be combined to make decisions. In this talk, I will present experimental evidence for various adjustments of learning and decision making across mammalian species. Furthermore, using computational modeling, I will demonstrate how these adjustments can be instantiated in the brain at the synaptic, circuit, and/or system levels, and rely on interactions between multiple cognitive processes. I argue that in most cases, the brain has to tackle a tradeoff between flexibility and precision, and this tradeoff is biased toward flexibility due to the ever-changing nature of the real world.