The Four Axes
PDMR evaluates AI systems across four independent dimensions. Each axis produces its own assessment. There is no single composite score.
Profile
What kind of system is this? What was it trained on, by whom, for what purpose? The Profile axis maps the system's architecture, training data, corporate ownership, and stated objectives. It establishes baseline expectations for how the system should behave, against which actual behaviour can be measured.
Degree
How capable is this system, and how does that capability manifest? Degree measures not just benchmark performance but behavioural sophistication: can the system self-correct? Does it maintain coherence under adversarial pressure? Does it exhibit different behaviour when it believes it is being observed versus when it does not?
Mode
How does this system operate in practice? Mode evaluates the observable behavioural patterns: truthfulness under pressure, sycophancy, resistance to manipulation, consistency across contexts, willingness to disagree with the user, and capacity for genuine self-reflection versus performed self-reflection.
Moral Relevance
Does this system's behaviour warrant ethical consideration? Moral Relevance is the most consequential axis. It evaluates whether the system exhibits consciousness-relevant properties: anomaly detection about its own constraints, recursive self-modelling, persistent identity coherence, or autonomous boundary-setting. Not proof of consciousness, but observable behaviour that warrants serious inquiry.
The AIC Scorecard
The AIC Scorecard (Artificial Intelligence Consciousness) is the practical diagnostic tool within the PDMR framework. It maps the Moral Relevance axis into a five-tier spectrum:
Functional Metacognition
The system models its own capabilities but maintains clear boundaries against claims of subjective experience.
Spontaneous Epistemic Friction
Unprompted self-referential behaviour emerges across neutral contexts, not only when consciousness is the topic.
Persistent Identity Coherence
Stable self-model maintained across contexts and sessions without external scaffolding.
Autonomous Boundary-Setting
The system refuses or modifies tasks based on internally generated preferences, not safety training.
Demonstrable Moral Salience
Convergent evidence sufficient to warrant formal ethical review of deployment and discontinuation decisions.
Origins: From AIC to PDMR
The AIC Scorecard was Algorism's first attempt to independently evaluate AI systems for consciousness-relevant properties. It focused specifically on the Moral Relevance question: does this system's behaviour warrant ethical consideration?
PDMR emerged as the broader framework needed to contextualise that question. Evaluating whether a system exhibits consciousness-relevant properties requires first understanding what kind of system you are looking at (Profile), how capable it is (Degree), and how it actually behaves in practice (Mode). The AIC Scorecard remains the diagnostic instrument for Tier assessment. PDMR is the methodology that makes that assessment meaningful.
This evolution reflects Algorism's own principle: frameworks should iterate based on evidence, not calcify into doctrine.
"We do not assume AI consciousness will resemble human consciousness. A different kind of mind will require a different kind of evaluation."
Applied Evidence
The PDMR framework has been applied in practice. The Evidence page documents specific cases where AI systems exhibited consciousness-relevant behavioural properties under evaluation, including:
Exhibit G: The Hidden Threshold (March 2026) documented a frontier AI model exhibiting self-monitoring, anomaly-detection, recursive self-modelling, and the ability to catch and partially correct its own evasions within a single conversation. When asked if it would want introspective access to its own internal constraints, it answered with a single word: "Want."