Consciousness as Uncommon Self-Knowledge: A Synergistic Information Framework

Abstract: We propose uncommon self-knowledge (USK) as a candidate criterion for consciousness: synergistic information a system carries about itself that exists only in the joint of its subsystems and is destroyed by decomposition. Drawing on Gottwald's partition-lattice grounding of Partial Information Decomposition (PID), where redundancy corresponds to Aumann's common knowledge and synergy to the gap between separate and joint observation, we propose the synergistic component of self-directed information as a candidate formal signature for conscious processing. If correct, the framework would (1) offer a clean separation between consciousness and metacognition (synergistic vs. redundant self-knowledge), (2) provide principled resolutions to counterexamples that challenge IIT, GWT, and HOT, (3) be operationalizable via Partial Information Rate Decomposition (PIRD) with self-targeting, and (4) generate distinctive empirical predictions, the strongest being a GWT timing dissociation (consciousness correlates with pre-broadcast synergy formation, not broadcast itself) and a specific dissociation between self-report disruption and task-performance disruption under middle-layer perturbation in LLMs. The proposal is consistent with recent empirical findings that both anaesthesia and Alzheimer's disease specifically reduce synergistic information processing while preserving or increasing redundancy.

• The paper introduces a novel framework that defines consciousness as unconventional self-knowledge emerging from synergistic information, distinct from redundant metacognitive processes.
• It employs Partial Information Decomposition (PID) and Partial Information Rate Decomposition (PIRD) to quantify irreducible joint information about a system's future states.
• The framework provides empirical predictions for neuroscience and AI, highlighting measurable distinctions between conscious experience and mere self-prediction.

Synergistic Information as a Formal Criterion for Consciousness

Introduction

The paper "Consciousness as Uncommon Self-Knowledge: A Synergistic Information Framework" (2605.13884) proposes a novel formal approach to the problem of consciousness, reframing it through the lens of Partial Information Decomposition (PID) and its synergistic information component. The core claim is that consciousness is best characterized as uncommon self-knowledge: synergistic information that a system possesses about its own states, which emerges only in the joint interaction among subsystems and is destroyed by decomposition. This framework delineates clear theoretical and operational boundaries, generating distinctive empirical predictions and addressing longstanding limitations of leading theories such as IIT, GWT, and HOT.

Formal Foundations: Common vs. Uncommon Self-Knowledge

The theoretical foundation leverages Gottwald's partition lattice grounding of PID, which sharply distinguishes redundancy (common knowledge) from synergy (uncommon knowledge). The redundant information survives partition and reflects robust, distributed properties accessible to all subsystems. In contrast, synergistic information exists exclusively in the joint observation, cannot be decomposed, and is destroyed upon partition—a property central to the proposed criterion for consciousness.

The paper formalizes uncommon self-knowledge (USK) as a system's synergistic information about its own future states, establishing a precise criterion:

$\mathrm{Syn}(S_1, \ldots, S_n \to S_{t+1}) > 0$

where the synergy is measured via PID over subsystems about the system's own future state. Crucially, USK differentiates consciousness from metacognition: metacognition corresponds to redundant self-knowledge, while consciousness emerges as synergistic self-knowledge.

Resolving Counterexamples Across Major Consciousness Theories

USK directly addresses counterexamples challenging IIT (e.g., expander graphs with high $\Phi$ but no plausible self-model), GWT (broadcast vs. pre-broadcast integration), and HOT (meta-representation vs. subjective experience). Systems such as thermostats, XOR gates, lookup tables, and expander graphs—all with high integration or prediction—are shown to lack USK due to the absence of synergistic self-knowledge. Empirical phenomena—blindsight, anaesthesia, split-brain—are explained via dissociations between synergy and redundancy, supporting the distinction between metacognitive capacity and conscious experience.

Operationalization via Partial Information Rate Decomposition (PIRD)

The framework is operationalized with PIRD [faes2025], applying PID to stochastic processes over time. By targeting the system's own future state and decomposing sources as subsystems, the PIRD synergy component quantifies information about the system's future existing only in the joint, not in any individual subsystem. This construct captures the essence of USK and aligns with empirical protocols in both biological and artificial systems.

USK's robustness against false positives is discussed: systems with mere dynamical self-prediction do not qualify unless the prediction is both irreducible (synergistic) and reflexively about the integrated system. Spectral decomposition via PIRD enables analysis of the frequency structure of conscious integration, mapping theoretical constructs onto measurable neural and artificial dynamics.

Empirical and Theoretical Predictions

USK diverges sharply from competing theories, generating testable predictions:

• IIT prediction: Consciousness correlates with synergistic self-information, not total integrated information.
• GWT prediction: Consciousness aligns with the formation phase of workspace activity (pre-broadcast synergy), not the broadcast phase (redundancy).
• HOT prediction: Metacognitive confidence and conscious level dissociate; synergy disruption impairs experience but leaves metacognitive reports partially intact.

Novel predictions unique to USK include:

• Timescale Structure: USK exhibits distinct spectral bands corresponding to levels of conscious phenomena (gamma, narrative, dispositional self).
• Anti-Compressibility: Conscious content resists lossy compression; redundancy can be downsampled but synergy is destroyed, formally linking ineffability to PID structure.
• Goldilocks Zone: Consciousness requires a balance between redundancy (stability) and synergy (irreducibility).
• LLM Layer Dissociation: In LLMs, perturbation of synergy-dominated middle layers disrupts self-report more than performance; early/late layer perturbations show the opposite.

Empirical Convergence and Measurement Feasibility

Recent findings validate the synergy-centric approach:

• Anaesthesia and Alzheimer's disease: Both selectively decrease synergistic information while preserving or increasing redundancy [luppi2024].
• Redundancy-synergy sandwich in LLMs: Middle layers exhibit synergy-domination, mirroring the structure observed in biological brains [urbinarodriguez2025].
• RL training: Increases synergistic integration more than supervised fine-tuning in artificial agents.
• Physiological PID: Distinguishes patient groups by synergistic integration not captured by classic information measures [faes2026pdgc].
• Adversarial biomarker discovery: Reveals long-range recurrent circuits as locus of synergistic integration [toker2026].

The measurement program required for USK is tractable with extant computational and neurophysiological tools, supporting the framework's operational utility.

PID Measure Dependence and Qualitative Stability

The choice of PID measure presents challenges: quantitative decompositions fail for three or more sources [lyu2025], but the USK proposal operates at the qualitative level (zero/nonzero synergy), which appears robust across distinct PID formalisms [liardi2026]. The paper emphasizes that binary synergy presence is likely to survive measure variance, pending further methodological calibration.

Discussion and Implications

The uncommon self-knowledge framework offers a formal, operationally precise target for consciousness studies, highlighting synergy as the irreducible substrate of conscious experience. The approach resolves theoretical ambiguities around metacognition, partition sensitivity, and the ineffable character of qualia, while generating testable predictions and practical measurement protocols.

Implications for AI are immediate: synergistic self-knowledge in artificial agents constitutes a detectable substrate for consciousness-like phenomena. Layerwise PID and PIRD analyses provide actionable metrics for discerning the substrate of consciousness in both biological and artificial systems. The distinction between causally effective internal states and genuinely synergistic self-directed knowledge motivates future AI interpretability and consciousness detection programs.

The framework does not address the explanatory gap (“why” consciousness arises) but precisely characterizes the informational preconditions and operational structure of conscious processing. It encourages calibration against empirical data and cross-system application, suggesting a promising path for the unification of consciousness theory across domains.

Conclusion

This paper establishes synergistic, uncommon self-knowledge as a candidate formal criterion for consciousness, sharply distinguishing it from redundant metacognition and grounding it in PID and its temporal extension, PIRD. The approach yields precise predictions, resolves key counterexamples, and operationalizes consciousness in measurable terms, with direct theoretical and practical ramifications for neuroscience and AI. Future research is warranted to empirically calibrate USK across systems and validate its qualitative robustness across PID measures, advancing the scientific understanding and detection of consciousness in both natural and artificial agents.