The Impact of Long-term Memory in the Iterated Prisoner’s Dilemma

Long term memory stores critical data for long periods of time. In this paper we show that the addition of long term memory grants agents playing the iterated prisoner’s dilemma (IPD) a competitive advantage. This paper extends previous work [1] in which IPD players utilized a limited subset of previous game actions to drive gameplay. We introduce a novel representation of strategies for the IPD which consist of a pair of finite state machines- one which records long term memory into a set of scalar variables, and a second which drives actions based on the memory data. Control agents act in much the same way but do not have access to long term memories.