Two papers I wrote with Roland Mühlenbernd last winter. We mainly explored what happens when the players are able to invent new messages in signaling games.
The important finding in the first paper is that the players ability to make up new messages overcomes signaling bottlenecks and leads to successfull communication even if the number of strategies is high. When coupled with a reasonable mechanism for message extinction, the agents will play perfect strategies in the end.
The second paper explores the evolution of the vowel space with this technique. The resulting partitions of the vowel space resemble these of actual languages.