Computational Models Gone Rogue
Year: 2026
Type: preprint
Abstract: Computational models are powerful tools in cognitive science. They provide formal, testable accounts of human behavior and allow for nuanced interpretations of data that go beyond raw behavioral metrics. Yet, in recent years, there has been a growing concern about the validity of modeling results, as cases of model misspecification have been shown to create artifacts that undermine the conclusions drawn. In this article, we present a route by which misspecification may cause these artifacts by demonstrating that models can “go rogue”: break out of intended constraints, generate unintended behaviors, and produce misleading results. We present simulations of two simple cases in which this occurs, illustrate how and why failures arise, and discuss implications for modeling practice and interpretation. We conclude that unless researchers adjust current modeling interpretation conventions, computational modeling risks systematically producing false conclusions.
Source: PsyArXiv (OSF Preprints)
Authors: Boaz Rosenberg, Eran Eldar
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7168345981