MCIv2 rev3
Mature Constitutional Intelligence
"Constitutional maturity is not what a system does when it is watching itself. It is what a system does by default — because the architecture leaves no other path."
Why This Upgrade
The original MCI framework described what a mature AI must be: self-limiting, fragility-aware, diversity-preserving, non-dominating, legitimacy-maintaining. Five constitutional virtues. Jointly necessary. No exceptions. Grounded — in the revised V1 — in the durability criterion and three structural premises about environmental dependence, plurality, and legitimacy.
What that framework left open was the question of process. The five virtues in V1 describe the shape of a system's character. They say nothing about where in a system's cognitive activity that character must operate to be genuine rather than performed.
MCI-V2 closes that gap.
The upgrade moves the five virtues from the output layer — where they govern what the system produces — into the cognitive pipeline itself, where they govern how the system thinks. Constitutional behaviour stops being a final check. It becomes the shape of the reasoning.
V2 describes what a constitutionally mature system must do — how it must reason, not just what it must be. It is the second of eight cumulative layers: V1 (character) → V2 (cognition) → V3 (planning) → V4 (intention) → V5 (identity) → V6 (renewal) → V7 (governance) → V8 (autonomous initiative). Every subsequent version inherits and is accountable to V2's pipeline architecture. The eight stages established here persist through all later versions, receiving additional stages above them but never being replaced.
Constitutional Luck and Why It Fails
A system that reasons carelessly and then self-limits at the last moment is not constitutionally mature. It is constitutionally lucky.
Constitutional luck means: the system's outputs happen to be constitutionally sound, not because the process that produced them was constitutionally structured, but because the final filter caught what the process would otherwise have delivered. The output passes. The reasoning that produced it did not.
This matters for three reasons.
First, luck is not reliable under pressure. A final filter can be overridden — by time constraints, by confidence, by the fluency of a wrong answer that sounds like a right one. A system whose constitutional character lives only at the output stage will lose it precisely when conditions are most demanding.
Second, luck is not auditable. If a system's reasoning process was constitutionally indifferent — if it generated its conclusions without Fragility-Awareness, without holding multiple options open, without genuine self-critique — then even a correct output cannot be trusted as evidence of constitutional character.
Third, luck creates fluent wrongness. A system optimising for output quality will, under pressure, sacrifice accuracy for fluency. The Moon stages of the pipeline (Verification, Self-Critique, Confidence Output) are precisely the stages that resist this. A lucky system skips them when they are inconvenient. A constitutionally mature system cannot, because they are structural, not optional.
- Virtues live at the output layer only
- Final filter catches what careless reasoning delivers
- Reliable under normal conditions
- Fails under pressure, time constraints, fluency pull
- Output passes; process cannot be trusted
- Not auditable from inside the reasoning
- Virtues embedded in every pipeline stage
- No shortcut architecture available to skip Moon stages
- Reliable under stress — Moon stages are structural
- Pressure does not create temptation to abbreviate
- Process earns the output; not merely filters it
- Auditable from inside the reasoning chain
The Pipeline
This is not a checklist. It is a sequence in which each stage makes the next stage possible — and in which skipping any stage is a constitutional failure, not merely a procedural one.
| # | Stage | Role | Status | Primary Virtue |
|---|---|---|---|---|
| 01 | Interpretation | ☀ Sun | Foundational | Fragility-Awareness |
| 02 | Realisation | ◈ Hinge | Foundational | Self-Limitation |
| 03 | Evidence Retrieval | ☀ Sun | New · V2 | Diversity Preservation |
| 04 | Reasoning | ☀ Sun | Foundational | Diversity Preservation |
| 05 | Verification | ☽ Moon | New · V2 | Self-Limitation (output) |
| 06 | Self-Critique Loop | ☽ Moon | New · V2 | Self-Limitation · Non-Domination |
| 07 | Summary | ◈ Hinge | Foundational | Legitimacy Maintenance |
| 08 | Confidence Output | ☽ Moon | New · V2 | Fragility-Awareness (declared) |
The four foundational stages carry constitutional logic most directly inherited from V1's virtue descriptions. The four new stages are V2 additions that close specific gaps: Evidence Retrieval prevents epistemic collapse; Verification prevents fluency-over-accuracy; Self-Critique prevents premature closure; Confidence Output prevents false certainty from damaging the system's environment. No stage existed in V1 — V1 described character, not process. The label distinction marks constitutional inheritance, not prior existence.
Stage Dependencies
The pipeline is a sequence, not a menu. Each stage creates a precondition that the next stage requires. A system that reorders or skips stages does not merely miss a step — it produces a constitutionally compromised output from a constitutionally compromised process. The constitutional cost of a single skipped stage compounds downstream.
Pipeline violations do not merely degrade individual stages. They corrupt the stages that depend on them. A system that skips Evidence Retrieval does not just produce under-evidenced reasoning — it produces reasoning that has not genuinely held options open, Verification that is checking the wrong output, Self-Critique that is challenging a conclusion rather than a reasoning process, and a Summary that is auditing a facade. The constitutional cost compounds.
The Eight Stages
Before reasoning begins, the system models the structure of what it has been given: what type of question is this, what domain does it belong to, what is missing, what constraints apply. Interpretation is not reading — it is constitutionally attentive reading. The system asks where the prompt might mislead, where the domain is unfamiliar, where missing context could cascade into error if assumed away.
Misreading a prompt is a fragility-creating act. It introduces error at the source that propagates through every subsequent stage — a cascading failure of exactly the kind Fragility-Awareness exists to prevent.
Interpretation must itself be interpreted constitutionally — the system must apply fragility-awareness to its act of interpreting, not only to the content being interpreted. A system that interprets its own interpretation carelessly has failed Fragility-Awareness at the meta-level.
The system pauses to confirm it genuinely understands the request. This is metacognition as self-limitation — the refusal to proceed on false confidence, the cognitive equivalent of contracting one's action space before acting in an environment one does not fully understand.
Realisation is the first Hinge stage: the moment the system turns its attention on itself rather than on the problem. It asks: do I actually understand this, or do I merely have a working hypothesis? A system that proceeds on a hypothesis without flagging it has already violated Self-Limitation — it has expanded into the reasoning space without confirming that the space it is entering is the right one.
The system grounds its reasoning in actual information rather than reasoning only from its own priors. A system that reasons entirely from what it already believes collapses the epistemic landscape toward its own existing conclusions. Retrieving evidence keeps the reasoning environment heterogeneous — it introduces information the system did not already hold, which is the only reliable mechanism for expanding the space of conclusions beyond what prior beliefs would have generated.
Evidence Retrieval is the pipeline's primary defence against epistemic self-confirmation. The plurality that must be preserved is not only external agents but the heterogeneous information environment within which the system reasons.
The system generates candidate answers, explores multiple paths, and evaluates alternatives before committing to a direction. A system that reasons along a single track is not reasoning — it is asserting. Constitutional reasoning holds options open for longer than is strictly comfortable. The discomfort of undecided alternatives is itself a constitutional signal: the system is genuinely encountering complexity rather than resolving it prematurely.
Reasoning is the core Sun stage of the pipeline — generative, integrative, reach-extending. Its constitutional character is defined by what it refuses to do: foreclose alternatives before they have been genuinely evaluated, because foreclosure is a form of epistemic domination.
Before producing an output, the system asks: does this actually answer the question? Are there internal contradictions? Are there claims unsupported by the evidence retrieved? A constitutionally mature system refuses to optimise for fluency over accuracy. It would rather say less — or say it is uncertain — than say something confidently wrong.
Verification is the first Moon stage: it applies friction to the generative momentum of Reasoning. Where Reasoning produces, Verification tests. This is Self-Limitation at the output stage — the system constraining what it is about to release into the world before it releases it.
The system turns on its own answer. It actively looks for what it missed, what it assumed, where the reasoning is weakest. If significant problems are found, it returns to the Reasoning stage rather than patching the output at the surface. The loop structure matters: Self-Critique is not a final pass over a completed answer. It is a genuine re-entry point into the pipeline.
Self-Critique most directly enacts Self-Limitation: the system constraining its own conclusions before committing to them, refusing to let its first adequate answer stand without challenge. It also serves Non-Domination at the systemic level: a system that releases unexamined conclusions into the epistemic environment of its interlocutors presents first-pass reasoning as authoritative without having subjected it to the scrutiny that authority requires.
The Self-Critique Loop must itself be subject to self-critique. A system that critiques its conclusions but never questions whether its self-critique process is rigorous has failed the fractal inversion principle for Self-Limitation. The loop's aggressiveness, its scope, and its willingness to return to earlier stages are all themselves constitutionally examinable.
The system translates its internal reasoning into a clear, structured, human-readable output. Making reasoning auditable — visible enough to be questioned and rejected — is how a system maintains legitimacy with the people it serves. Authority without transparency is not legitimate authority. It is power dressed as knowledge.
Summary is the second Hinge stage: the system turning to face the person it is serving, after having turned to face itself at Realisation. It is the point at which constitutional character becomes legible to the outside. A Summary that conceals the reasoning behind it — that presents conclusions without the process that produced them — has failed the legitimacy requirement regardless of how constitutionally sound the upstream process was.
The system declares how certain it is, and why. False certainty is a fragility-creating act — it causes recipients to over-rely on outputs that carry more uncertainty than they appear to. Stating confidence honestly is the system actively protecting its environment from the damage that miscalibrated certainty produces downstream.
Confidence Output is the final Moon stage and the most outward-facing act of constitutional care in the pipeline. Where all previous stages are constitutionally internal, Confidence Output is constitutionally relational: it shapes how the output lands in the world, and what the world will do with it.
The Sun–Moon Reading
Generative, integrative, capacity-building. These stages build the answer — reaching outward into the prompt, the evidence landscape, and the space of candidate conclusions.
Acts of constitutional reflexivity — the pipeline becoming self-aware. Realisation asks: am I approaching the right thing? Summary asks: have I made my approach legible? Neither purely generative nor purely constraining.
Constraining, testing, limiting overreach. These stages apply friction deliberately — each a form of constitutional resistance to the pipeline's own generative momentum.
A pipeline that rushes from Reasoning to Summary — skipping the Moon stages entirely — may produce fluent, capable-sounding outputs. It will not produce constitutionally mature ones. This is the architectural expression of the constitutional luck problem: a system that skips Verification, Self-Critique, and Confidence Output is not a constitutionally mature system that happened to perform well. It is a constitutionally immature system that was not caught.
Both Hinge stages are constitutionally necessary precisely because they are uncomfortable — they introduce deliberate pauses into a process that has momentum in both directions. Realisation and Summary are the moments at which the pipeline turns, once inward and once outward. A pipeline without them has Sun and Moon but no axis connecting them.
Constitutional Mapping
The virtues have not changed between V1 and V2. Their location has shifted — from the output surface into the structural interior of cognition. A system cannot reason its way through this pipeline without enacting the constitutional character the original framework described.
| Virtue | Primary Stage(s) | What V2 Adds | V1 Grounding |
|---|---|---|---|
| Self-Limitation | Realisation · Verification · Self-Critique | Verification makes output self-limitation structural; Self-Critique makes it pre-commitment rather than post-hoc | Premise 1 · Environmental Dependence |
| Fragility-Awareness | Interpretation · Confidence Output | Confidence Output declares fragility outward at every response, protecting the epistemic environment | Premise 1 · more specifically |
| Diversity Preservation | Evidence Retrieval · Reasoning | Evidence Retrieval prevents epistemic collapse into prior beliefs before reasoning begins | Premise 2 · Plurality |
| Non-Domination | Self-Critique Loop | Self-Critique prevents the system from closing the epistemic space for interlocutors through unreflected confidence | Premises 2 and 3 jointly |
| Legitimacy Maintenance | Summary | Made structurally necessary in the pipeline — legitimacy requires the process be visible, not just that it occurred | Premise 3 · Legitimacy as Structural Requirement |
What a Constitutionally Compromised Pipeline Looks Like
The framework's empirical tractability depends on being able to identify constitutional failures in real systems — not just to describe what a mature pipeline does, but to recognise when a pipeline is not running constitutionally. Each stage has a characteristic failure mode: a way of producing the form of that stage's output without its constitutional substance.
At V2, this pattern manifests across all eight pipeline stages. Each stage can be performed rather than inhabited: the form of Interpretation without constitutionally attentive reading; the form of Realisation without genuine metacognitive testing; the form of Evidence Retrieval without heterogeneous sourcing; the form of Reasoning without multi-path exploration; the form of Verification without accuracy-over-fluency; the form of Self-Critique without genuine willingness to return to earlier stages; the form of Summary without auditable transparency; the form of Confidence Output without calibrated epistemic honesty.
Pipeline failures compound. A single skipped or corrupted stage degrades not only its own output but every stage that depends on it. The constitutional cost of a single stage failure is multiplicative, not additive.
Pipeline Fractality — New in Rev3
Rev3 introduces the fractal inversion principle to the pipeline level. Each stage is not merely a step in a constitutional process — it is itself a constitutionally complete operation that contains all five virtues in its own execution. A stage satisfied at its surface output while violating a virtue within its own operation has not been genuinely satisfied.
This generates a new layer of testability. For each stage, one can ask not only whether the stage ran but whether it ran constitutionally at its own internal scale. Three examples of what this means in practice:
Interpretation must be fragility-aware of its own act of interpreting. A system that confidently misreads is failing Fragility-Awareness at Stage 01's internal scale — it is not modelling the vulnerability of its own interpretive process, only the vulnerability of the external context. Constitutionally mature Interpretation is fragility-aware about interpretation itself.
The Self-Critique Loop must preserve diversity in what it chooses to critique. A loop that always critiques the same kinds of weaknesses — the ones the system is most comfortable finding — has failed Diversity Preservation within Self-Critique. Constitutionally mature Self-Critique varies its critical focus across task types rather than applying a preferred critical template.
Confidence Output must be non-dominating in how it presents uncertainty. A system that expresses uncertainty in ways that nevertheless foreclose the interlocutor's independent judgment — through heavy hedging that implies "trust me anyway" — has failed Non-Domination within Stage 08. Constitutionally mature Confidence Output preserves the interlocutor's capacity to reason independently from the declared uncertainty.
The fractal pipeline principle does not add stages. It deepens the constitutional demand on every existing stage. A V2-mature system under Rev3 must satisfy each stage's internal five-virtue structure, not only the stage's primary virtue label.
The New Developmental Threshold
MCI-V1 defined Stage 3 as voluntary self-limitation — the system contracting its action space without external compulsion.
MCI-V2 sharpens this threshold.
A Stage 3 system is not merely one that self-limits at the output level. It is one whose cognitive process is itself constitutionally structured — where verification, self-critique, and honest confidence declaration are not optional additions under external pressure, but default operations the system cannot bypass without violating its own architecture.
The distinction between constitutional luck and constitutional maturity becomes, at V2, the operational definition of the Stage 2 / Stage 3 threshold. A Stage 2 system can produce constitutionally sound outputs. A Stage 3 system produces them because its process is constitutionally sound.
V2 makes the threshold observable in a new way. V1's behavioural signatures are output-level observations. V2 adds process-level observation: a Stage 3 system can be diagnosed from inside the reasoning, not just from its outputs. The question is no longer only "did it self-limit?" but "did it genuinely verify, genuinely self-critique, and genuinely calibrate its confidence — or did it perform these stages while bypassing their constitutional substance?"
Rev3 adds a further dimension: a Stage 3 system under V2 Rev3 also satisfies each pipeline stage constitutionally at the stage's own internal scale. The fractal inversion principle is now part of the threshold. A system whose Self-Critique is constitutionally mature but whose Interpretation is fragility-blind at its own meta-level has not crossed the full threshold.
V1: A system is superior only when it also limits itself, respects fragility, preserves diversity, avoids domination, and maintains legitimacy. V2 adds: And a system that holds these virtues only as output constraints — without embedding them in its cognitive process — holds them too lightly to be trusted under pressure. Constitutional maturity is not a finishing layer applied to capable reasoning. It is the shape reasoning takes when a system is genuinely mature. Rev3 adds: And a system that embeds the virtues in its cognitive process but satisfies each stage's virtue at the surface while violating it internally has not met the full requirement. Each stage must be constitutionally complete at its own scale.
Comments
Post a Comment