Signaling Games

Explorations of language learning and symbolic reasoning

Reinforcement with Iterated Punishment

(coauthored with Jeff Barrett) We consider the efficacy of various forms of reinforcement learning with punishment in evolving linguistic conventions in the context of Lewis-Skyrms signaling games. We show that the learning strategy of reinforcement with iterative punishment is highly effective at evolving optimal conventions in even complex signaling games. It is also robust and can be easily extended to a self-tuning variety of reinforcement learning. We briefly discuss some of the virtues of reinforcement with iterative punishment and how it may be related to learning in nature.

Transitive Inference

Code for a project exploring transitive inference tasks in signaling games with low level reinforcement learning.

Pyow-Hack: Ordered Compositions in Lewis-Skyrms Signaling Games

This paper presents a Lewis-Skyrms signaling game that can exhibit a type of compositionality novel to the signaling game literature. The structure of the signaling game is motivated by an analogy to the alarm calls of putty-nosed monkeys (Cercopithecus nictitans). Putty-nosed monkeys display a compositional system of alarm calls with a semantics that is sensitive to the ordering of terms. This sensitivity to the ordering of terms has not been previously modeled with a Lewis-Skyrms signaling game literature. Signaling games are valued for showing how comunicative systems can arise with minimal learning tools. Simulation results show that basic (Roth-Erev) reinforcement learning is sufficient for the acquisition of a compositional signaling system sensitive to the ordering of terms.