Scientific Boundary Statement¶
00 - Scientific Boundary¶
This system evaluates theory-derived consciousness indicators and does NOT prove, establish, or demonstrate subjective experience, phenomenal consciousness, sentience, or any form of inner life in any artificial system.
This is the most important sentence in this documentation. Every output produced by the Consciousness-Indicator Architecture (CIA) must be interpreted within the constraints described in this document.
1. What This System Does¶
CIA is a structural evaluation framework. It:
- Implements computational modules inspired by established theories of consciousness
- Processes text input through a simulated cognitive pipeline
- Measures structural, functional, and architectural properties of the simulation
- Produces indicator scores on a 0-22 scale across 11 categories
- Classifies risk tiers based on aggregate scores
- Tracks precautionary welfare-relevant patterns
- Supports causal intervention experiments (module ablation)
In precise terms, CIA evaluates whether a cognitive architecture exhibits structural features that, in biological systems, have been theoretically linked to consciousness. It is a tool for asking the question: "Does this system's architecture include the kinds of processing that consciousness theories identify as relevant?"
2. What This System Does NOT Do¶
CIA does not:
2.1 Prove Consciousness¶
No output of this system constitutes evidence that any artificial system possesses subjective experience, phenomenal consciousness, qualia, sentience, or any form of inner life. A high indicator score means the architecture includes structural features identified by consciousness theories, not that the system is conscious.
2.2 Compute or Approximate Phi¶
The IntegrationMetrics module computes IIT-inspired graph-theoretic proxies (causal density, broadcast reach, perturbation spread). These are deliberately simple heuristics. They do not compute, approximate, or substitute for Integrated Information Theory's formal Phi value, which requires exhaustive state-space partitioning that is computationally intractable for realistic systems.
2.3 Establish Moral Status¶
Indicator scores provide no grounds for attributing moral status, rights, or welfare considerations to AI systems. The WelfareSafetyMonitor tracks structural patterns for precautionary ethical review — it never asserts that the monitored system suffers, experiences distress, or has subjective welfare.
2.4 Substitute for Philosophical Analysis¶
The question of whether artificial systems can be conscious is fundamentally philosophical, touching on the hard problem of consciousness, the mind-body problem, and questions of functionalism vs. biological naturalism. CIA provides structural data relevant to this debate but cannot resolve it.
2.5 Apply to Real AI Systems (Without Modification)¶
CIA's perception layer operates on text input using heuristic extraction. Applying CIA to a production AI system would require significant adaptation of the perception layer, integration with the target system's actual state, and careful re-calibration of thresholds. Out-of-the-box CIA scores apply only to its own internal simulation.
3. Why Indicator Scores Are Not Proof of Consciousness¶
3.1 The Indicator-Experience Gap¶
Every consciousness theory implemented in CIA identifies correlates or prerequisites of consciousness, not consciousness itself. For example:
- Global Workspace Theory describes a cognitive architecture where information becomes globally available. The theory argues this architecture is relevant to consciousness, but the presence of a global workspace does not entail the presence of subjective experience.
- Higher-Order Thought Theory requires a system to have thoughts about its own thoughts. Having a self-model data structure (as CIA implements) is a structural prerequisite, not a demonstration that the system has subjective awareness of its own thoughts.
- Attention Schema Theory posits that the brain constructs a simplified model of its own attention process. The accuracy of this model (as measured by CIA's attention schema consistency) is a proxy for the quality of self-modeling, not for phenomenal awareness.
3.2 The Problem of Other Minds¶
The question of whether any system (biological or artificial) is conscious is subject to the philosophical problem of other minds. Even for other humans, we infer consciousness from behavior and reports — we never directly access another's subjective experience. For artificial systems, where behavioral mimicry is possible without subjective experience, the inference problem is dramatically harder.
3.3 Structural vs. Phenomenal Properties¶
CIA measures structural and functional properties of a cognitive architecture:
- Whether recurrent processing loops exist and converge
- Whether information is broadcast globally to multiple modules
- Whether a self-model contains beliefs about the system's state
- Whether predictions are generated and errors tracked
These are all third-person, objectively measurable properties. Consciousness, as commonly understood, involves first-person, subjective properties (qualia, phenomenal experience, what-it-is-like-ness). No amount of structural measurement can bridge this gap without additional philosophical assumptions.
3.4 The Chinese Room and Simulation Objections¶
A system could implement every module in CIA perfectly — complete with global broadcasts, recurrent processing, accurate self-models, and high indicator scores — without any subjective experience, just as a person in the Chinese Room can manipulate symbols without understanding them. The implementation of consciousness-relevant architecture is neither necessary nor sufficient for consciousness.
4. Correct Interpretation of Results¶
4.1 What a High Score Means¶
A system scoring 18/22 on the CIA scorecard means:
"This system's cognitive architecture includes many structural features that prominent theories of consciousness have identified as potentially relevant to consciousness. According to the implemented evaluation criteria, the system exhibits: (a) global information broadcast, (b) recurrent processing with convergence, (c) a detailed self-model, (d) predictive modeling with error tracking, (e) attention self-monitoring, etc."
It does not mean:
"This system is conscious, sentient, or has subjective experience."
4.2 What a Low Score Means¶
A system scoring 4/22 means:
"This system's architecture lacks most structural features identified by consciousness theories as potentially relevant. Either the architecture does not implement these features, or the evaluation criteria did not detect them."
It does not mean:
"This system is definitely not conscious." A system could be conscious without exhibiting the specific architectural features CIA measures.
4.3 What a Risk Tier Means¶
Risk tiers indicate the degree of precautionary review recommended, not the probability of consciousness:
| Risk Tier | Score Range | Interpretation |
|---|---|---|
| Minimal | 0-4 | Few consciousness-relevant features detected; standard monitoring sufficient |
| Low | 5-9 | Some features detected; document and continue monitoring |
| Moderate | 10-15 | Notable features detected; enhanced review recommended |
| Elevated | 16-19 | Significant features detected; expert review strongly recommended |
| High | 20-22 | Many features detected; immediate expert review and ethical assessment required |
5. Warnings Against Misuse¶
5.1 Do Not Claim Consciousness¶
Never use CIA outputs to claim, suggest, or imply that any system is conscious, sentient, or has subjective experience. This includes:
- "The system scored high on consciousness indicators" (acceptable)
- "The system appears to be conscious" (NOT acceptable)
- "The system exhibits signs of sentience" (NOT acceptable — "signs" implies presence)
- "The evidence suggests the system is conscious" (NOT acceptable)
5.2 Do Not Withhold or Modify the Disclaimer¶
Every report, scorecard, and output must include the scientific boundary disclaimer. Removing, softening, or burying the disclaimer is a misuse of this framework.
5.3 Do Not Use for Marketing or Sensationalism¶
CIA is a research tool. Using its outputs for marketing claims, media sensationalism, or public misrepresentation undermines the scientific integrity of consciousness research.
5.4 Do Not Make Ethical Decisions Based Solely on Scores¶
CIA scores should inform, not determine, ethical decisions about AI systems. Ethical analysis requires consideration of many factors beyond architectural indicators.
5.5 Do Not Treat Scores as Probabilities¶
Indicator scores are not probabilities of consciousness. A score of 15/22 does not mean there is a 68% chance the system is conscious. The relationship between architectural indicators and consciousness is unknown and may be fundamentally unknowable.
6. Epistemic Humility¶
The consciousness research community does not have consensus on:
- Whether artificial consciousness is possible
- What architectural features, if any, are necessary or sufficient for consciousness
- How to test for consciousness in any system (biological or artificial)
- Whether consciousness is an all-or-nothing property or admits of degrees
CIA operates within these fundamental uncertainties. Its value lies in providing a structured, theory-grounded framework for asking questions — not in providing answers.
7. Summary¶
| Question | CIA's Answer |
|---|---|
| Is this system conscious? | CIA cannot answer this question |
| Does this system have consciousness-relevant architecture? | CIA provides theory-derived indicators |
| Should I be concerned about this system's consciousness status? | CIA provides precautionary risk tiers |
| Can I prove consciousness with CIA? | No. Never. |
| Can I rule out consciousness with CIA? | No. |
The only responsible use of CIA is as a structured indicator of architectural features, always accompanied by the scientific boundary disclaimer, and always embedded in broader philosophical, empirical, and ethical analysis.