Coupled Control, Structured Memory, and Verifiable Action in Agentic AI (SCRAT -- Stochastic Control with Retrieval and Auditable Trajectories): A Comparative Perspective from Squirrel Locomotion and Scatter-Hoarding
2026-04-03 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors compare intelligent AI systems to squirrels because squirrels naturally combine the need to act, remember, and check their surroundings despite not having full information all the time. They study different kinds of squirrels and create a model that includes memory, decision-making, and verification processes working together. From this, the authors propose hypotheses about how fast feedback, organized memory, and verification inside decision loops can make AI more reliable. They also suggest that splitting AI roles into different parts might reduce errors when information is uneven. Overall, the authors provide a new way to test how control, memory, and verification can work together in AI.
Agentic AIPartial observabilityControl systemsEpisodic memoryLatent dynamicsObserver-belief stateVerificationScatter-hoardingHierarchical control modelRole differentiation
Authors
Maximiliano Armesto, Christophe Kolb
Abstract
Agentic AI is increasingly judged not by fluent output alone but by whether it can act, remember, and verify under partial observability, delay, and strategic observation. Existing research often studies these demands separately: robotics emphasizes control, retrieval systems emphasize memory, and alignment or assurance work emphasizes checking and oversight. This article argues that squirrel ecology offers a sharp comparative case because arboreal locomotion, scatter-hoarding, and audience-sensitive caching couple all three demands in one organism. We synthesize evidence from fox, eastern gray, and, in one field comparison, red squirrels, and impose an explicit inference ladder: empirical observation, minimal computational inference, and AI design conjecture. We introduce a minimal hierarchical partially observed control model with latent dynamics, structured episodic memory, observer-belief state, option-level actions, and delayed verifier signals. This motivates three hypotheses: (H1) fast local feedback plus predictive compensation improves robustness under hidden dynamics shifts; (H2) memory organized for future control improves delayed retrieval under cue conflict and load; and (H3) verifiers and observer models inside the action-memory loop reduce silent failure and information leakage while remaining vulnerable to misspecification. A downstream conjecture is that role-differentiated proposer/executor/checker/adversary systems may reduce correlated error under asymmetric information and verification burden. The contribution is a comparative perspective and benchmark agenda: a disciplined program of falsifiable claims about the coupling of control, memory, and verifiable action.