A Symbiotic Architecture for the Emergence of Artificial Consciousness: A Testable Triadic Framework

Summary

A proposed modular cognitive architecture called the Symbiotic Triadic Architecture (STA) is designed to exhibit functional hallmarks of conscious-like processing without assuming subjective experience. It integrates three subsystems: a neuro-symbolic reasoning module for explainable action selection, a hypothesis management subsystem using Bayesian neural networks for perceptual interpretations under uncertainty, and a narrative episodic memory module for self-modeling and metacognitive calibration. These components are coordinated through a Global Workspace Bus with ethical gating and cryptographic provenance. The paper maps the architecture to existing computational paradigms, identifies five integration challenges with mitigation strategies, and outlines a three-stage implementation roadmap from 2026 to 2031+. It aims to operationalize conditions for studying conscious-like functional properties rather than claiming consciousness.

Study at a glance

Abstract

The scientific study of artificial consciousness has long been polarized between ontologically ambitioustheories that resist empirical test and purely instrumental architectures that sidestep questions of self-modeling, accountability, and interpretability. This paper presents a proposed modular cognitivearchitecture—the Symbiotic Triadic Architecture (STA)—designed to instantiate functional hallmarkscommonly associated with conscious-like processing, without presupposing the existence of subjectivephenomenal experience.The architecture integrates three dynamically coupled subsystems: a neuro-symbolic reasoning module(HT) responsible for explainable action selection and linguistic report via retrieval-augmented generationand symbolic rule engines; a hypothesis management subsystem (HQ) that maintains competingperceptual interpretations under epistemic uncertainty through Bayesian neural networks and activeinference loops; and a narrative episodic memory module (EN) supporting self-modeling, metacognitivecalibration, and affectively tagged temporal knowledge graphs. These components are coordinated through a Global Workspace Bus (GWB) implemented as an auditable message-passing middleware withpolicy-enforced ethical gating and cryptographically signed provenance metadata.We provide a component-by-component mapping to existing computational paradigms, demonstratingarchitectural feasibility at Technology Readiness Levels 4–6. We identify five critical integrationchallenges—including real-time latency coordination, semantic interoperability across heterogeneousrepresentations, and the absence of standardized evaluation benchmarks for self-narrative coherence—andpropose concrete mitigation strategies for each. A three-stage implementation roadmap definesmeasurable milestones from simulated environments (2026–2027) through embodied prototyping(2027–2029) to socially situated deployment (2029–2031+), with explicit pass/fail criteria at each stage.This work constitutes Part II of a two-paper research program. A companion paper establishes thetheoretical foundations of the STA within Global Workspace Theory, neurophenomenology, andoperational criteria for functional consciousness. The present paper addresses the complementaryengineering question: can this architecture be built, tested, and audited with current technology? Ratherthan claiming the realization of consciousness, we aim to operationalize the conditions under whichconscious-like functional properties can be systematically studied, compared, and held accountable.