Bayesian Hierarchical Modeling of Attentional Resource Allocation Under Dual-Task Interference: A Latent Variable Framework for Decomposing Cognitive Load Trajectories

Abstract

Contemporary dual-task paradigms inadequately capture the temporal dynamics of attentional resource depletion, limiting mechanistic inference in cognitive load research. The present study introduces a Bayesian hierarchical latent variable framework capable of decomposing individual-level cognitive load trajectories from response time distributions and pupillometric indices simultaneously. Using a mixed-cohort sample (N = 214) across three dual-task interference conditions, we calibrated posterior predictive distributions via Markov Chain Monte Carlo sampling to isolate task-switching costs, bottleneck latencies, and residual executive control variance. Model comparison via Widely Applicable Information Criterion demonstrated superior fit relative to conventional ANOVA-based approaches. Results indicate that attentional resource allocation follows a non-stationary stochastic process modulated by working memory capacity and prior task exposure. These findings provide a methodologically rigorous platform for individualized cognitive load profiling in clinical and applied behavioral settings.