A parameter recovery assessment of a wide class of evidence accumulation models of decision-making

Computational modeling has become indispensable in investigating the dynamics of decision making processes. A prominent category of models in this domain are Evidence Accumulation Models (EAMs), which model both the decisions people make and the time they take. Many variations have been proposed which modify the drift rate, diffusion rate, and decision thresholds, encoding increasingly complex dynamics into the EAM framework. However, adding model features complicates parameter recovery, making model interpretation more difficult. In this work, we perform a parameter recovery study to a variety of common binary choice EAMs, identify the specific challenges for each, and explore how to improve their parameter recoverability. Though previous studies have addressed this question, they have been piecemeal in nature, with different groups applying different computational methods to study different models. We aim to unify this body of literature using the best currently available computational methods. Further, we present the first, to our knowledge, Bayesian analysis of diffusion conflict models. Our purpose here is to be thorough, not exhaustive or comprehensive. With this in mind, this article catalogues a number of results, some previously shown and some new. Further, it illustrates different approaches to model analysis. This article is intended to be a resource for researchers interested in utilizing EAMs for studying decision-making processes, providing insights into the challenges associated these models, how to analyze them in light of those challenges, and examples of how to address those challenges.