Tune in for the live stream on YouTube or Twitter.
Abstract: Central questions in religious studies are increasingly approached through statistical modeling of large-scale cross-cultural data. Findings, however, have been conflicting, in part due to the challenges of missing data, expert disagreement, and confounding. Inspired by the successful application of the (inverse) Ising model to social and biological systems, we show how inferring a “landscape model” from these data can provide a new lens to investigate cultural evolution; the Bayesian nature of the estimation process allows us to handle missing data, biased coverage, and expert (dis)agreement in a rigorous fashion, while providing a new framework for thinking about the underlying causal processes. Landscape models not only provide new insight into the fundamental constraints that drive religions into a subset of observed forms; they also provide a holistic view of how these constraints combine to produce a complex landscape with well-defined peaks and cultural floodplains, and allow us to make quantitative predictions about features of ancient religions that are unknown to the field. By going beyond simple regression models, they enable us to hypothesize in new ways about hidden variables, causality, and dynamical evolution.