Coarse Grained Models of Cultural Transmission

Project Description

Coarse grained models of cultural transmission are those which describe transmission with lower granularity than individual-based or intra-population models, summarizing the results of social learning over larger structures, or aggregated over spans of time instead of making point observations. A coarse grained model may also introduce observation units which are aggregates of the information actually passed by individuals during social learning. Coarse grained models will allow us to make contact with, and explain, phenomena which are not at the same scale as our detailed individual-level models of cultural transmission and social learning.

The classic modeling work by Boyd and Richerson, or Cavalli-Sforza and Feldman, is comparatively "fine-grained," in that models describe stable equilibria achieved in comparatively short time intervals, with observations made upon individuals and their traits. The observed traits, in a fine-grained model, are the same units of information which are copied and imitated by individuals. Such models follow the structural framework of theoretical population genetics, or more rarely, epidemiology. Their analysis typically leads to an understanding of "ecological time" dynamics, and the equilibria of the random process(es) describing the model.

Fine grained models form the basis for nearly all archaeological discussion of CT today, despite the fact that archaeologists deal with a highly aggregated, often "time averaged" record of past artifactual discard and human behavioral traces. Several decades of formation process studies, taphonomic studies of a variety of materials and artifact classes, and a growing appreciation of the limits that the sedimentary record places upon the questions we can ask (i.e., "time perspectivism") should be informing our efforts to adapt fine grained cultural transmission models to archaeological use, but to date little or no attention has been given to bridging the gap between "ecological" or "human" time, and the spatiotemporal scales over which archaeological evidence of human culture can inform us.

This problem is identical to that faced by paleobiologists who seek to provide evolutionary explanations of the fossil record, although the massive temporal aggregation characteristic of that record has largely led to a bifurcation between "micro" and "macroevolutionary" questions and models. Archaeological evidence (construed broadly), on the other hand, is capable of informing us about cultural change on time scales ranging from years to millennia, and thus we need to consider a continuum of relations between "micro" and "macro," rather than simply bifurcating our questions and theories.

Our problem is thus structurally similar to scaling or "renormalization" problems in physics, where microscopic descriptions of phenomena often need to be aggregated and understood at a variety of observational scales. Processes critical to understanding observations at high energies (and thus, sub-atomic scales) are invisible when a system is viewed at the molecular and atomic scales (and thus, lower energies), and have only statistical effects when a system is viewed at a bulk or macroscopic scale (e.g., a ferromagnetic sample of iron). At each successive scale, we construct a "coarse grained" description of the system by summarizing and "averaging over" detail at the lower level.

I am engaged in a project to study the properties of common cultural transmission models as we "coarse grain" them in various ways:

  • temporal aggregation
  • spatial or population aggregation
  • classificatory aggregation of observation units
  • all of the above, since this is what we see in in the empirical record.


This project forms the basis of my Ph.D. dissertation research at the University of Washington. The results of these researches will be released in the form of 3-4 published or in press papers, along with a dissertation document outlining the background for the research and research methods.

One paper, an initial examination of temporal aggregation or "time averaging" upon the classic neutral model adapted by Neiman from Sewall Wright, Motoo Kimura, and others, has been submitted to the _Journal of Anthropological Archaeology_, and was presented at the 2012 meeting of the Society for American Archaeology. [PREPRINT]

Project Notes

10 Apr 2016 Research Priorities for 2016
22 Mar 2016 Next Steps for Classifying Seriations to Temporal Network Models
22 Feb 2016 Limits of model resolution for seriation classification
16 Feb 2016 Feature Engineering for Seriation Classification
14 Feb 2016 Identifying Metapopulation Models from Simulated CT and Seriations
26 Jan 2016 Loss Functions for ABC Model Selection with Seriation Graphs as Data
24 Jan 2016 Amazon EC2 AMI for deep neural networks and classification problems
01 Sep 2015 IDSS Seriation Software Version 2.3 released
27 Jul 2015 Performance of simuPOP in SeriationCT Across Platforms
14 May 2015 SeriationCT Sample Size Series experiments
27 Apr 2015 SeriationCT Next Steps
23 Feb 2015 Experiment Plan for seriationct-1
15 Feb 2015 CTMixtures Simulation in Docker
07 Dec 2014 CTMixtures Analysis -- Equifinality-4 Progress and Tasks
04 Dec 2014 CTMixtures Analysis -- Splitting the Analysis
28 Nov 2014 Implementing temporal networks in Python, Part 2
20 Nov 2014 CTMixtures Analysis -- RF to GBM and Next Steps
07 Nov 2014 CTMixtures Analysis -- Equifinality 3 Notes
06 Oct 2014 CTMixtures Equifinality - Calibration Experiment and Conclusions
22 Sep 2014 CTMixtures Equifinality - Calibration Issues and Solutions
18 Sep 2014 CTMixtures Equifinality Analysis - Initial Ideas
17 Sep 2014 CTMixtures Equifinality Data Export
16 Sep 2014 CTMixtures Status for Batch Equifinality-1
15 Sep 2014 CTMixtures Cost Analysis
14 Sep 2014 CTMixtures Experiment Configuration and Execution
14 Sep 2014 CTMixtures Cluster Setup and Computing Environment
31 Aug 2014 SeriationCT Experiment Outline
10 Aug 2014 Observables and Parameters in CTMixtures
10 Aug 2014 CTMixtures Experiment Design
28 Jul 2014 Implementing temporal networks in Python
13 Jul 2014 Temporal Networks in SeriationCT
09 Jul 2014 Time Averaging and the CT Mixture Model
17 Jun 2014 Description of the SeriationCT Model
14 Jun 2014 Open Problems in Coarse Grained CT Theory
24 May 2014 Experiment Planning for CT Mixture Model
26 Apr 2014 Convergence of CT Rule Mixtures
02 Apr 2014 TODO for Semantic Axelrod project
18 Mar 2014 Coarse Graining for Structured Information
03 Nov 2013 Classification Experiment - Exploratory Data Analysis
26 Sep 2013 Classification Experiment - Data Reduction and Cleaning
27 Aug 2013 Classification Experiment Research Questions
09 Aug 2013 Classification Experiment Notes
29 Jul 2013 Classification Experiment Protocol
20 Jul 2013 Classifications in CTPy (Part 2) -- Modeling Dimensionality, Performance
13 Jul 2013 Classification in CTPy (Part 1)
02 Oct 2012 Coarse-Graining, History, and CT Models
01 Oct 2012 Experiment - Statistical Power of Slatkin Exact vs Conformism
22 Sep 2012 Slatkin Exact Test -- Error and Statistical Power
14 Jun 2012 Renormalization Theory and Cultural Transmission in Archaeology
23 Feb 2011 Renormalization and Archaeological Applications