Self-assembling Games

Using Logic to Evolve More Logic: Composing Logical Operators via Self-Assembly

“…Thus, the key idea is that rather than having to evolve a binary disposition from scratch, the agents may learn to compose two pre-evolved unary dispositions appropriately. In one way, this is an extension of previous work on self-assembly, insofar as the complex signalling behaviour examined here is the same as that which is examined in Barrett and Skyrms (2017)-namely, the evolution of logical operations; however, the mechanism by which this behaviour evolves is wholly different. 5 I compare the efficacy and efficiency of several means for evolving logical operations by examining the cumulative success rates of the agents' evolving strategies under 2 See also LaCroix (2018b).…”

“…All other things being equal, this is the simplest dynamic that readily illustrates the particular phenomena of interest. 4 Further, since this is the dynamic that Barrett and Skyrms (2017) discuss, using the same dynamic allows for more simple comparison of their results (as a baseline) with the results of the present models.…”

Truth and probability in evolutionary games

“…Simple games may evolve from the ritualization of the decisions of agents, and more complex games may evolve from the composition of simpler games (Barrett and Skyrms 2015). Here we briefly consider two composite Lewis-Skyrms signaling games (Lewis 1969) (Skyrms 2010).…”

“…That is, signaling games may evolve from the ritualization of the actions of very simple agents. 6 That basic forms of language may evolve from the interactions of such simple agents illustrates, among other things, that the preconditions for the evolution and use of language are significantly more modest than philosophers have typically thought. 7 And showing how primitive 4 There are other low-rationality types of learning like win-stay/lose-shift where the agents do not get stuck in such suboptimal pooling equilibria.…”

“…5 See Skyrms (2010) for an introduction to learning and population models and the relationship between them and Huttegger, Skyrms, Tarrès, and Wagner (2014) for a more detailed discussion of evolutionary dynamics for signaling games. 6 See Barrett and Skyrms (2015) for an account of how both simple and composite signaling games may evolve by the ritualization of agent actions. 7 Concretely, this illustrates that metaphysical notions such as ideal forms, innate ideas, pure concepts of the understanding, or an a priori language of thought are not required to explain how notions of truth and probability might coevolve with descriptive language helps to make the point that even such traditional philosophical language may simply evolve to provide agents with whatever expressive power is in fact required for successful coordinated action given the task at hand.…”