The role of forgetting in the evolution and learning of language

Lewis signaling games illustrate how language might evolve from random behavior. The probability of evolving an optimal signaling language is, in part, a function of what learning strategy the agents use. Here we investigate three learning strategies, each of which allows agents to forget old experience. In each case, we find that forgetting increases the probability of evolving an optimal language. It does this by making it less likely that past partial success will continue to reinforce suboptimal practice. The learning strategies considered here show how forgetting past experience can promote learning in the context of games with suboptimal equilibria. We often decry our own forgetfulness, wishing that we could remember more of the past in order to successfully guide our actions today. It is tacitly believed by many that forgetfulness is a human frailty, which should be reduced wherever possible. The present study may shed doubt on this widely held belief. In this paper, we present a model of language evolution, where forgetfulness plays an important role. In this model, a learning rule which remembers the entire past, basic Herrnstein reinforcement learning, fairs much worse than three other learning rules that discard varying amounts of past experience. These results show how forgetting can be a virtue in the context of games with suboptimal equilibria.