Learning without Recall: A Case for Log-Linear Learning
This provides a tractable framework for modeling rational agents in networks, addressing a foundational problem in social learning theory.
The paper tackles the complexity of Bayesian learning in social networks by proposing a 'Learning without Recall' model, which simplifies inference by assuming agents do not recall past observations or reason about others' beliefs, and analyzes how time-varying priors affect learning rates.
We analyze a model of learning and belief formation in networks in which agents follow Bayes rule yet they do not recall their history of past observations and cannot reason about how other agents' beliefs are formed. They do so by making rational inferences about their observations which include a sequence of independent and identically distributed private signals as well as the beliefs of their neighboring agents at each time. Fully rational agents would successively apply Bayes rule to the entire history of observations. This leads to forebodingly complex inferences due to lack of knowledge about the global network structure that causes those observations. To address these complexities, we consider a Learning without Recall model, which in addition to providing a tractable framework for analyzing the behavior of rational agents in social networks, can also provide a behavioral foundation for the variety of non-Bayesian update rules in the literature. We present the implications of various choices for time-varying priors of such agents and how this choice affects learning and its rate.