Theory in Spring
Evolutionary Theory Across Scales of Organisation Part II
- Date: Apr 8, 2021
- Time: 11:00
- Speaker: Anne Kandler (MPI for Anthropology)
- Location: virtual platform
- Host: Chaitanya Gokhale
One of the major challenges in cultural evolution is to understand why and how various forms of social learning are used in human populations, both now and in the past. To answer this question directly social learning processes would need to be observed directly so that fine-grained individual-level data detailing who learns from whom can be generated. But outside of controlled experimental conditions, large longitudinal datasets of this kind are difficult to obtain, especially in historical or anthropological contexts. This is not to say that no such data exist, but in many case studies of interest the available data are in the form of frequencies of different variants of a cultural trait in the population at one or several points in time. As these frequency data often present the only direct empirical information about past cultural traditions and the forces affecting them, researchers have attempted to use the population-level patterns to infer learning biases that may underlie them. But given the large variety of learning biases that have been identified in the literature and other factors that may affect the frequency data we ask in this talk how accurately this inverse problem can be addressed.
In the first part we focus on the coarse distinction between unbiased social learning and biased social learning, based on data of low and high temporal resolution. We show that in a number of circumstances population-level frequency patterns generated by an unbiased learning (drift) process may not conform to neutral expectations solely due to unmodelled properties of the cultural system. In more detail, we show that using statistics blind to (i) characteristics of the population such as its age structure or (ii) details of the learning process such as the learning of packages of cultural variants vs. the learning of single variants may generate misleading inference results. Further, we demonstrate that the quality of the data, in particular their completeness, can be of crucial importance. The presence, or absence, of rare variants as well as the spread behaviour of innovations may carry a stronger signature about underlying processes than the dynamic of high-frequency variants and the consistency between empirical data and hypotheses about social learning processes can depend entirely on the completeness of the data set.
In the second part we advocate the use of the generative inference approach to simultaneously evaluate the consistency of a number of learning hypotheses with the available data while also accounting for - potentially complex - demographic and cultural properties of the cultural system. This approach consists of a generative model that establishes a causal link between learning processes and observable frequency data that then are evaluated for statistical consistency. Besides identifying the most likely learning process given the data, this framework determines the breadth of processes that could have produced these data equally well, which in turn allows us to quantify the level of equifinality of the inverse problem and to evaluate the limits of inferring social learning processes from population-level data.