12. Social Learning in a Model of Adverse Selection

Social network learning: Uncertainty, heterogeneity, and the application in principal–agent relationships

“…In such a case we can regard a strategy with a high payoff as at best a "probably good" strategy. This paper draws on the social evolutionary learning algorithm introduced by [28], which assumes that an individual can learn with certain probabilities, which are determined by the payoffs obtained by her neighbors and herself (using roulette wheel selection):…”

“…In such a case we can regard a strategy with a high payoff as at best a “probably good” strategy. This paper draws on the social evolutionary learning algorithm introduced by [28], which assumes that an individual can learn with certain probabilities, which are determined by the payoffs obtained by her neighbors and herself (using roulette wheel selection): $$$$ {\mathrm{\mathbb{P}}}_{p,i\Rightarrow k}=\frac{e^{\theta {\pi}_{p,k}}}{\sum \limits_{j\in \mathbf{nb}(i)}{e}^{\theta {\pi}_{p,j}}}, $$$$ where $$$ {\mathrm{\mathbb{P}}}_{p,i\Rightarrow k} $$$ is the probability that individual $$$ i $$$ learns from $$$ k $$$ at $$$ p $$$, and $$$ \theta $$$ is a parameter controlling the relative fitness weights. As would be the case with LBN, the neighbors' payoffs can influence an individual's belief.…”