Explaining temporally clustered errors with an autocorrelated Drift Diffusion Model

bioRxiv (Cold Spring Harbor Laboratory)

Abstract

Abstract A widely used framework for studying the computational mechanisms of decision-making is the Drift Diffusion Model (DDM). To account for fast and slow errors the DDM incorporates across-trial variability in parameters such as the drift rate and the starting point. Although these variability parameters enable the model to reproduce both fast and slow errors, they rely on the assumption that parameters are independently sampled over trials. As a result, the DDM effectively predicts that errors—whether fast or slow—occur randomly in time. Here, we show that in empirical data this assumption is violated, as error responses are temporally clustered. To address this limitation in the model, we introduce the autocorrelated DDM, in which drift rate, starting point, and boundary evolve from trial-to-trial according to first-order autoregressive (AR1) processes. Using simulations, we demonstrate that, unlike the across-trial variability DDM, the autocorrelated DDM naturally accounts for temporal clustering of errors. We further show that model parameters can be reliably recovered using Amortized Bayesian Inference, even with as few as 500 trials. Finally, in a pre-registered analysis we find that the autocorrelated DDM provides the best account of error clustering across 8 separate datasets, including both human and mouse data. These results establish that accounting for temporal fluctuations in computational parameters is essential for accurately characterizing decision-making behavior.

Authors 4

1. Robin Vloeberghs corresponding

   KU Leuven

   Affiliation as printed

   Brain & Cognition, KU Leuven, Belgium

2. Francis Tuerlinckx

   KU Leuven

   Affiliation as printed

   Quantitative Psychology and Individual Differences, KU Leuven, Belgium

3. Anne E Urai Aachen Cognitive Psychology

   Leiden University

   Affiliation as printed

   Cognitive Psychology, Leiden University, The Netherlands

4. Kobe Desender

   KU Leuven

   Affiliation as printed

   Brain & Cognition, KU Leuven, Belgium