Clarify why and how current AI agentic reasoning approaches work relative to biological reasoning

Determine the mechanisms by which existing AI agentic reasoning approaches operate and establish how these mechanisms compare to biological reasoning grounded in neural processes such as hierarchical cognition, multimodal integration, and dynamic interactions.

Background

The paper motivates a neuroscience-grounded framework for agentic reasoning and notes a gap between current AI methods and biological reasoning. Despite rapid progress in approaches such as LLM-based reasoning, VLMs, and reinforcement learning, the causal and computational underpinnings of their effectiveness remain insufficiently explained relative to well-studied neural mechanisms in humans.

Resolving this uncertainty would help align AI agentic reasoning with cognitive processes observed in the brain, potentially improving robustness, adaptability, and generalization of intelligent agents.