Abstract
This thesis attempts to study interaction and learning from a unified perspective. It is divided into three parts. Part I looks at deterministic dynamics applied to multidimensional beliefs. Part II develops random discrete learning and has the main mathematical results, generalizing the earlier benchmark model. In part III, a continuous version of social dynamics with noise is studied. In parts II and III, uncertainty in learning means examining probabilistic aspects of social dynamics. New models in the Degroot framework of social learning are introduced and developed. The mathematical tools range from dynamical systems to aspects of probability theory and stochastic calculus. Underlying the whole thesis is a strong current of thought on interaction and learning.