|Created: 11 Aug 2013||Modified: 23 Jul 2020||BibTeX Entry||RIS Citation|
Buchanan and Collard are organizing a symposium at next year’s SAA conference, and sent Carl and I both invites. I suggested doing a conference version of the “Semantic Axelrod” model that I’d proposed for the Mesoudi Springer volume on CT and the Paleolithic record. This is tangential to my main dissertation work, but it’s a long-standing research interest.
The list of research questions in my 2007 note are what led to thinking about semantic version of the Axelrod model. Take the following as an example:
“Worldview Similarity Bias”: One experiment we could run would be to select a focal individual each tick in a Moran process, and then examine the set of candidate models, finding the model(s) which have the most number of traits in common which share the same set of semantic relationships, and select one of their traits randomly to copy. Do we end up with persistent associations of concepts? How does the turnover rate of traits change? What level of variation do we see in the population is it lower, higher, or the same as we’d see in a pure neutral model? We’re essentially doing unbiased selection of traits from a biased selection of individuals.
First, note that “model” in the above refers to a “cultural model” – i.e., a neighboring agent.
But the overall model description is simply copying predicated on homophily, where the latter is measured on semantic relationships between the agent pair’s traits, rather than simple trait identity. This begs many questions, of course: what semantic relationships, between what types of concepts? What is “close enough”? But it’s recognizably the Axelrod model, even though I didn’t make the connection in 2007.
The Axelrod model seems to be one of the simplest models in which we can examine the micro-level causes of differentiation (or conversely, the causes of consensus or uniformity). In its original form, the dynamics are governed entirely by copying and a homophily rule which serves as a threshold for when copying is allowed.
There are a vast number of potential semantic relationships between concepts we could model. Which ones are important in answering archaeological questions?
In studying the evolution of technologies, especially over the long time spans characterized by Paleolithic deposits, we are often interested in explaining patterns in the richness, diversity, and use context of toolkits. In particular, we tend to contrast the toolkits of the Middle Paleolithic as low in diversity and class richness, as compared to the increasingly diverse and specialized toolkits of the Upper Paleolithic.
Much of the research into this transition in toolkit diversity focuses upon factors like adaptation and ecological change, or cognitive or biological differences among different groups of early Homo sapiens. (NEED TO REVIEW AND VERIFY THE CURRENT LIST OF EXPLANATIONS FOR THIS). But there is some recent work which seeks to determine whether such differences are a byproduct of evolutionary processes operating in an era of demographic change (REF - see 2012 Kandler and Steele introduction).
Certainly it is the case that a denser population will serve as a larger reservoir for cultural variation; with the right balance between innovation, population size, and the density of migration and social network contacts a set of cultural traits can find “dynamic refugia” and circulate through a metapopulation, avoiding the extinction which would be certain in a panmictic population.
And it’s clearly the case that a sparse metapopulation can lead to the formation of distinctive regions of cultural traits, and that cognitive biases such as homophily or conformist social learning can accentuate this “regionalization” of culture. But the toolkits we see are not random assortments of traits, even in the presence of randomizing connections such as migrations and trade connections.
Toolkits have functional cohesion, and are cumulative traditions of engineering responses to tasks given raw materials.