Toward a Principled Bayesian Workflow in Semantics: A Tutorial

When researchers study semantic phenomena using Bayesian cognitive models, they must validate these models to ensure they accurately formalize the underlying theory and predictions. This is particularly important, as Bayesian models have gained increasing popularity in semantics inrecent years, yet the question of how to validate these models has not been addressed in the lit-erature. Here, we present an R and Stan-based tutorial on five model validation methods basedon the principled Bayesian workflow in cognitive science (Gelman et al., 2020; Betancourt, 2018;Schad et al., 2021). We apply these methods to a cognitive model on semantic representations ofnatural function words. Specifically, we demonstrate how to implement and interpret (1) prior pre-dictive checks, (2) computational faithfulness, (3) model sensitivity, (4) parameter recovery, and(5) posterior predictive checks. The tutorial includes path diagrams for an easier understandingof the methods, troubleshooting checklists to address undesired outcomes, and R and Stan codeto make it easier to apply these methods to new situations.