Simulating Language 2015, lecture 4 pre-reading

We want you to read Oliphant (1996), which I think is a lovely, short, simple paper on the evolution of communication. The author, Mike Oliphant, is no longer in academia, but in the late 90s he worked in the LEC (Language Evolution and Computation research unit – thats the group I am part of, and the name you see emblazoned all over this page) as a post-doctoral researcher and co-supervised my undergraduate dissertation on a similar topic.

You should be equipped to understand this paper without further preamble from me, based on the recent lectures and labs, but there are a couple of points where a little clarification might help:

1. Oliphant uses a different representation of signalling systems from us. Whereas we are representing agents as matrices of association strengths between meanings and signals, and then using the winner-take-all procedure to model production and reception, Oliphant encodes send and receive behaviours direct in his agents’ genotypes: so the genotype 1010 corresponds to something like the Lewis-style signalling system 1->b, 2->a for production, and b->1, a->2 for reception. Oliphant’s representation is simpler, but the matrix-based system we are using is more extensible (as you will find out in week 4).
2. In Section 3.2 of the paper, Oliphant starts talking about the Prisoner’s Dilemma. This is a standard problem in game theory, and there are nice online introductions: when you get to that point, you could do worse than look at the Wikipedia page for the Prisoner’s Dilemma (no need to read past the end of the section titled “Strategy for the iterated prisoner’s dilemma”), and then have a shot at this neat little online tutorial.
3. Oliphant talks a bit about “drift” when explaining the meandering trends you see in e.g. Figure 8 in his paper. He is talking about genetic drift, which arises when have heritable variation but no natural selection, i.e. when all variants have the same fitness. In this scenario, the frequencies of variants fluctuate randomly – sometimes a variant might increase in frequency, then it might decrease, and these changes are driven purely by chance, by individuals being “lucky” and being selected disproportionately often to reproduce, or “unlucky”, being sampled a bit less than you’d expect.

Once you have completed the reading, do the usual post-reading quiz, and then have a look at how I would have answered.

References