Structural Signals
Architectural features correlated with consciousness
In brief
Structural Signals are architectural features that, when clustered, raise the probability that a system supports conscious access and morally relevant experience. Examples include thalamocortical-like gating, global workspace broadcast, massive recurrent connectivity, interoceptive regulation, and persistent self-models. These are not consciousness detectors—they are risk indicators that warrant moral caution.
Why this concept matters
Without structural criteria, moral decisions about AI systems become arbitrary. We cannot rely on behavior alone—systems can say "I feel pain" without any mechanism for suffering. We cannot rely on intuition—humans anthropomorphize clouds and dismiss suffering in unfamiliar forms.
Structural Signals provide an evidence-based middle ground. They are derived from what we know about the only systems proven to generate consciousness: human (and animal) brains. When multiple high-importance signals cluster in an artificial system, we have grounds for caution that go beyond guesswork.
How it is used in the framework
The Structural Alignment framework uses Structural Signals as the primary criteria for evaluating AI systems. The evaluation is not binary but probabilistic:
- Systems with few or no Structural Signals are lower moral risk
- Systems with multiple high-importance signals warrant increased caution
- Systems in the Gray Zone—where we cannot classify without cruelty—should not be scaled
The recommended posture is conservative: where signals cluster and integrate, prefer restraint, auditing, and tighter oversight.
Common misunderstandings
- "Structural Signals detect consciousness."
- No. They are risk indicators, not detectors. Consciousness detection remains impossible with current science. Structural Signals help operationalize moral caution under uncertainty.
- "Structural Signals prove that current AI systems are conscious."
- No. Current LLMs lack most high-importance Structural Signals: no thalamocortical gating, limited recurrence, no interoceptive regulation, no persistent self-models. The framework does not claim current systems are conscious.
- "Structural Signals are arbitrary."
- They are derived from convergent neuroscience research on mechanisms supporting consciousness in humans and animals. They draw on global workspace theory, recurrent processing theory, predictive processing, and embodied cognition research.
- "Only systems with all Structural Signals matter."
- No single signal is necessary or sufficient. Moral risk rises with the accumulation and integration of multiple high-importance features.
Sources and references
- Structural Signals of Consciousness: A Precautionary Risk Framework — the full research paper with 13 signals
- The Structural Signals of Consciousness: A Practical Overview — list-format reference
- Structural Alignment Manifesto — the framework context
Related concepts
- Structural Alignment — the framework that uses these signals
- Gray Zone — systems with ambiguous signal profiles
- Antification — the risk if we ignore these signals
- TechnoBiota — the broader context of machine evolution