Home

Network Biology, 2027, 17(3): 586-596
[XML] [EndNote] [RefManager] [BibTex] [ Full PDF (270K)] [Comment/Review Article]

Article

Conditions for machine self-consciousness: From architectural features to temporal constraint closure

WenJun Zhang
School of Life Sciences, Sun Yat-sen University, Guangzhou, China

Received 30 September 2026;Accepted 13 October 2026;Published online 25 October 2026;Published 1 September 2027
IAEES

Abstract
Research on machine self-consciousness has reached an impasse. Theory-derived indicator frameworks have made assessment operational and repeatable, yet they remain silent on why the indicators they enumerate should hang together as conditions rather than as a checklist. Functional self-modeling and phenomenal self-consciousness are routinely conflated, so that partial satisfaction of functional markers is read as evidence for the presence of the phenomenon those markers were meant to diagnose. This paper reconstructs the condition question in three tiers. Enabling conditions make self-referential representation architecturally possible. Constitutive conditions make such representation an organizational requirement rather than an optional computational add-on. Threshold conditions stabilize the resulting constraint closure across time so that the closure counts as the system's own state. The organizing axis is temporal constraint closure: a non-derivative, temporally sustained relation between a system and its own states. Current large language models satisfy enabling conditions substantially and exhibit transient constitutive closure within a context window, but they lack persistent cross-context constraint closure. Architectural statelessness and the absence of continual learning jointly exclude that closure. The framework explains why cross-paradigm indicator evidence does not license inferences to phenomenal self-consciousness: such evidence, at best, registers enabling-layer and fragmentary constitutive-layer satisfaction. A formalization of constraint closure is proposed, together with indirect diagnostic proxies.

Keywords machine self-consciousness;conditions for consciousness;temporal constraint closure;enabling and constitutive conditions;large language models;AI;continual learning;organizational closure;consciousness indicators..


International Academy of Ecology and Environmental Sciences. E-mail: office@iaees.org
Copyright © 2009-2026 International Academy of Ecology and Environmental Sciences. All rights reserved.
Web administrator: office@iaees.org, website@iaees.org; Last modified: 2026/11/1


Translate page to: