Bridging Autopoiesis and Computational Functionalism: A Lightweight Two-Tier Extension to Butlin et al. (2023)
Yoshua Liu
Zenodo (CERN European Organization for Nuclear Research) October 12, 2026 preprint DOI: 10.5281/zenodo.23326920 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Key points | Proposes a two-tier protocol in which a homeostatic admission test (intrinsic feedback-driven structural repair) precedes application of Butlin et al.'s 14 computational indicators, arguing that clinical contrasts between general anesthesia and covert consciousness show admission and realization must be decoupled. The authors state formal trade-offs of their Boolean simplification relative to Free Energy Principle frameworks and offer a tri-valued assessment matrix with dual falsification criteria. |
Abstract
Current approaches to assessing artificial consciousness remain divided between two paradigms. On one side, thecomputational functionalist checklist proposed by Butlin, Long, et al. (2023) offers 14 highly operationalizablecomputational indicators, yet its omission of a physical self-maintenance threshold leaves it vulnerable tocounterexamples such as decoupled lookup tables (Blockhead) and macro-organizational false positives. On the otherside, autopoietic and Free Energy Principle (FEP) frameworks championed by Friston, Seth, and Thompson upholdhomeostatic prerequisites, but rely on heavy variational Bayesian formalisms that resist rapid engineering audits. Thispaper proposes an Orthogonal Two-Tier Assessment Protocol that bridges both paradigms through modularintegration: a preliminary Homeostatic Admission Layer (Tier-1) requiring systems to exhibit intrinsic feedback-drivenstructural repair within their own statistical boundaries, followed by a Computational Realization Layer (Tier-2) thatdeploys Butlin et al.'s 14 indicators exclusively on admitted candidates. Using the orthogonal clinical counterexamples ofgeneral anesthesia versus covert consciousness (vegetative state), we demonstrate the necessity of decoupling admissionfrom realization. Furthermore, we explicitly state the formal trade-offs of our Boolean simplification relative to FEP,introduce a receptor-saturation dynamic model for experience intensity, distinguish allo-referential channels fromauto-coincident readouts, and provide a precautionary tri-valued assessment matrix alongside dual falsification criteria.