Testing early and late connectivity dynamics in the lab: an experiment using 4-agent micro-societies

How does the order of individuals' interactions affect the emergence of shared conventions at the population level? The answer to this question is relevant for a number of fields, such as cultural evolution, linguistics, cognitive science or behavioral economics. In this study we investigate experimentally how two different network connectivity dynamics affect the evolution of the diversity of cultural variants of the communication system. We report an experiment in the lab in which participants engage in a Pictionary-like graphical communication task as members of a 4-participant micro-society, interacting in pairs with the other three members of the community across 4 rounds. The experiment has two main goals: First, to evaluate the effect of two network connectivity dynamics (early and late) on the evolution of the convergence of micro-societies on shared communicative conventions under controlled conditions. Second, to compare the predictions of the agent-based model described in a previous study (Segovia-Martín, Walker, Fay, & Tamariz, 2019) against experimental data, and calibrate the model to find the best-fitting parameter setting. Our experimental data shows that, as predicted by the model, an early connectivity dynamic increases convergence and a late connectivity dynamic slows down convergence. We found significant differences between conditions in round 3 and round 4. We estimate the best-fit parameter combination for the 96 data structures coded. Medium to high content bias, neutral to egocentric coordination bias and memory size of 3 rounds was associated with a better model fit. In the light of the model evaluation and the experiment outcome, we discuss the impact of our predictions on social influence research and possible factors that might help to improve model precision.