Learning and Herding Using Case-Based Decisions With Local Interactions

We evaluate repeated decisions of individuals using a variant of the case-based decision theory (CBDT), where individuals base their decisions on their own past experience and the experience of neighboring individuals. Looking at a range of scenarios to determine the successful outcome of a decision, we find that for learning to occur, agents must have a sufficient number of neighbors to learn from and access to sufficiently independent information. If these conditions are not fulfilled, we can easily observe herding in cases where no best decision exists.