This work proposes the architecture for a unified framework which integrates three distinct types of expectations: Popperian, Skinnerian, and Gregorian. Popperian Expectations allow agents to predict the outcomes of actions from internal simulations without any physical risk. Skinnerian Expectations allow agents to gather real-world feedback from their environment to both update and refine existing expectations and update the logic of their internal models when variables in their environment change in unexpected ways. Lastly, Gregorian expectations can allow agents to reason about symbolic knowledge and social norms allowing for a higher abstraction of expectations. The architecture also introduces the Expectation Arbitration Engine (EAE) to manage the expectations from each module and determine the context in which the expectations should be applied.