A promising new foundation for conflict and control: Bridging the gap to verification. Comment on “Active inference and cognitive control: Balancing deliberation and habits through precision optimization” by Riccardo Proietti, Thomas Parr, Alessia Tessari, Karl Friston, & Giovanni Pezzulo

Visalli, Antonino · 2026 · Crossref

DOI: 10.1016/j.plrev.2025.11.008

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This commentary by Antonino Visalli addresses the fragmented theoretical landscape of cognitive control, specifically critiquing and evaluating a recent model by Proietti et al. that applies the Active Inference framework to unify deliberation and habits. The motivation for this review stems from persistent debates regarding how signals like conflict influence control allocation, with existing models proposing unintegrated functions ranging from conflict detection to effort coding. Visalli examines whether Proietti et al.’s proposal, which recasts control allocation as a hierarchical optimization problem balancing costs and benefits, offers a viable path toward theoretical unification and empirical verification. The commentary analyzes Proietti et al.’s model, which treats cognitive control as the precision over deliberative policies within a normative framework. Visalli highlights the model’s core assumption: that precision is dynamically adjusted based on the real-world success or failure of combined deliberative and habitual policies, rather than directly by conflict levels. To address a resulting "vicious circle" where policy failure reduces confidence in deliberation, the model introduces an "optimistic simulation" mechanism. Visalli argues that while this formulation is original, it lacks a mechanistic description of intra-trial temporal dynamics, specifically the decision-making processes that lead to policy selection. This omission is critical because experimental data in cognitive control rely heavily on reaction time (RT) distributions and their specific profiles, not just final choices. Visalli identifies a significant gap in the proposed framework: its inability to account for the heterogeneous temporal dynamics observed in classic conflict tasks such as the Stroop, Simon, and Flanker effects. He notes that sequential sampling models reveal distinct interference loci and dynamics, such as transient decay in Simon tasks versus sustained accumulation in Flanker tasks, which are reflected in different delta plot profiles. The commentary argues that without a bridge explaining how the model’s signals translate into these varied temporal dynamics, the framework remains disconnected from empirical evidence. Consequently, the model cannot yet explain how a single principle of precision optimization generates such diverse behavioral profiles, making it difficult to falsify against established data. The significance of this commentary lies in its call for rigorous empirical validation of the Active Inference approach to cognitive control. Visalli concludes that while Proietti et al.’s framework offers an elegant unification of pragmatic and epistemic signals, its greatest challenge is explaining the deep heterogeneity of decision-making dynamics. The future development of this theory depends on demonstrating how its unified principle can account for the rich, varied temporal profiles of conflict resolution observed in experimental settings. This critique underscores the necessity of integrating mechanistic descriptions of decision dynamics into high-level normative models to ensure they remain grounded in observable behavioral evidence.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• executive function