Opinion Shaping in Social Networks Using Reinforcement Learning

In this article, we consider a variant of the classical DeGroot model opinion propagation with random interactions, in which prescribed subset agents is amenable to control parameter. There are also some stubborn and that neither nor control. We map problem shortest path problem, where parameter coupled across controlled nodes because common resource constraint. Hence, not pure dynamic programming approach, reinforcement learning schemes for latter cannot be applied here maximizing average influence long run. view it instead as parametric optimization use nonclassical policy gradient scheme. analyze its performance theoretically through numerical experiments. situation when only certain interactions between observed.