Any arrangement of stabilized commitments operating in an environment that changes faster than the commitments do will eventually face a structural problem: how to acknowledge that operational practice has drifted from the commitments without dissolving the authority that grounds the commitments. This problem appears wherever the conditions hold, which is more places than is usually noticed. Religions, constitutions, scientific theories, philosophical positions, identities, marriages, employment relationships, legal systems, artificial intelligence training protocols, mathematical practice, professional standards, ideological movements — all are commitment systems. All face the same structural problem. All instantiate the same patterns of response.
This is a theoretical claim, not an empirical survey. The argument is that the structural pattern is real and the framework that analyzes it is portable across the domains where the pattern obtains. The empirical work of testing the framework in specific domains is the research program this paper aims to enable, not the work this paper performs.
Three primitives
A commitment system has three components.
A kernel is the stabilized arrangement of commitments that constitutes the system’s nominal core. This may be a text (a constitution, a foundational paper, a creed), a body of practice (the methodology of a craft tradition), a narrative (an identity’s organizing story), a relationship (the implicit terms a marriage rests on), or some combination. The kernel is what participants in the system point to when asked what the system is.
An authority structure is the mechanism that determines what counts as legitimate interpretation of the kernel. Authority structures may be external (a priesthood, a judiciary, an academic discipline, a board), distributed (the parties to a relationship, a community of practice, a peer-review system), internal (an individual’s self-interpretive practices), or composite. Authority structures have incentives, resource flows, and stakes in preserving their standing; treating them as politically neutral interpretive machinery misses what they are.
Drift is unmarked mutation of operational practice relative to the kernel. Drift is intrinsic because the environment the kernel was stabilized within continues to change after the kernel stabilizes. No commitment system preserves perfect fidelity to its kernel across time. The question is not whether drift occurs but whether the system’s authority structure can acknowledge drift without losing legitimacy.
The kernel is frame-relative. At any moment T1, the kernel is whatever participants in the system point to as the reference for that analysis. Drift between T1 and T2 is measured against the T1 kernel, not against some eternal fixed reference. The kernel itself can change between T1 and T2, and that change is itself drift to be acknowledged or denied. This treatment makes systems with evolving kernels analyzable in the same terms as systems with fixed ones.
The structural problem
The people with authority to officially acknowledge drift are usually the same people whose standing depends on pretending there isn’t any. A church hierarchy that admits doctrine has changed undermines its claim to unchanging truth. A constitutional court that admits it is making new law undermines its claim to just reading old law. A spouse who admits the relationship has fundamentally changed has to deal with what that means. An AI lab that admits its alignment framework has drifted has to address what the systems trained against the earlier framework actually do.
So most systems develop elaborate ways of not admitting drift — reinterpreting old commitments to mean new things, or quietly doing something different while insisting nothing has changed.
This is the framework’s central observation. The question is not whether drift happens. The question is whether the system can name it.
Five patterns of response
Systems handle drift in five characteristic patterns. The patterns are claims about attractor states in the configuration space of possible commitment system arrangements, not partitions of all possible systems. Real systems often mix patterns and many of the most interesting cases occur at boundaries.
Marked revision. The kernel is precisely specified. Authority is voluntary and grounded in expertise. Drift is formalized as revision: someone proposes a change, the change is checked, and if it works the system absorbs it. Acknowledgment is marked and legible rather than absorbed silently. Mathematics works this way. Some healthy long-term relationships work this way. Programming languages with formal proposal-and-deprecation procedures work this way. The pattern is stable when acknowledgment capacity matches environmental change rate.
Interpretive accretion. The text is fixed. Authority is grounded in continuity with the founding text. The formal mechanism for changing the text does not function or does not exist. Drift migrates entirely into interpretation. Everyone insists the kernel controls while the operational meaning shifts substantially. Brahmanical commentary on the Vedas works this way. Catholic doctrinal development works this way. Common law jurisprudence works this way. Constitutional originalism, despite its claim to fixity, works this way. The pattern is durable across millennia when the interpretive layer can absorb the operational drift the environment produces.
Diffuse reconstruction. The kernel is under-specified or intentionally ambiguous. No centralized authority structure exists to adjudicate. Many parties produce mutually incompatible readings claiming the same source. Platonism in philosophy works this way. The various interpretations of “alignment” or “safety” in AI development work this way. Terms like “freedom” or “justice” in contemporary political discourse work this way. The pattern persists indefinitely but lacks operational coherence. It often serves strategic purposes for parties who benefit from operational ambiguity.
Implicit practice. There is no codified kernel. The kernel is whatever the system does. Authority is grounded in practice itself. Drift is the mechanism rather than a failure of it. The UK constitution works this way. Craft traditions transmitting tacit knowledge through apprenticeship work this way. Long-term partnerships where the relationship is whatever the parties do together work this way. The pattern is stable as long as practice remains coherent. Breakdown is severe when practice loses coherence because there is no fixed referent to reconstruct from.
Anchored fixity. The kernel is formalized. The authority structure grounds its legitimacy in the kernel’s unchangeability and extracts substantial benefit from preventing kernel revision. Drift denial is the source of authority rather than a side effect. The pattern has two sub-types that produce radically different stability profiles.
Anchored fixity with an interpretive-accretion layer beneath the kernel can persist millennia. The Hindu Vedic-Brahmanical system, post-development Catholic doctrine, and the Confucian commentary tradition operate this way. The unrevisable kernel is paired with a substantial interpretive substructure that absorbs drift without surfacing kernel revision. The interpretive layer does the acknowledgment work the kernel itself cannot do.
Anchored fixity without an interpretive-accretion layer is structurally brittle. The kernel is supposed to govern operational practice directly. The Spartan Lycurgan system worked this way. Certain forms of religious fundamentalism work this way. Certain trauma responses that treat the traumatic event as unrevisable kernel without developing interpretive machinery to process it work this way. The pattern produces accumulating gap and catastrophic breakdown when environmental change exceeds what the kernel can govern. Spartan breakdown at Leuctra is the canonical case at civilizational scale; identity crises in fundamentalist deconstruction are the same pattern at personal scale.
The framework’s strongest cross-domain claim is that anchored fixity without an interpretive-accretion layer produces brittle-fixity failure identically across organizational, identity, and relational systems. Treatment approaches that work in one domain may transfer to others when the structural diagnosis is the same.
Why systems fail
Systems do not usually fail because participants are bad or stupid. They fail because the mechanism for admitting change is broken or captured by people who benefit from things staying officially the same.
The framework makes this explicit. Each pattern has its characteristic failure mode. Marked revision fails when environmental change accelerates beyond acknowledgment capacity, or when the authority structure develops extraction stakes in kernel preservation that did not exist before. Interpretive accretion fails when the interpretive layer cannot absorb the operational drift the environment produces, or when accretion becomes so distant from the kernel that participants notice the gap. Diffuse reconstruction fails by never resolving operational coherence; the failure is the persistent condition rather than an event. Implicit practice fails when practice loses coherence through generational discontinuity, environmental disruption, or authority structure breakdown. Anchored fixity without interpretive accretion fails catastrophically when environmental change exceeds kernel-governance capacity. Anchored fixity with interpretive accretion fails slowly through accumulated unrecognized drift that the interpretive layer can no longer plausibly absorb.
These are structural predictions, not contingent observations. The pattern plus the environmental rate plus the authority structure’s incentives produce the failure mode. The pattern is real because the failure mode follows from the structural conditions rather than from the specifics of any particular case.
Different positions see different things
Within any commitment system, participants at different positions see different things. Insiders, external observers, beneficiaries, victims, reformers, conservatives — each position produces a different classification of the same kernel-practice relationship. The disagreement has structure rather than being noise. Some positions are systematically unable to see drift they participate in producing, because seeing it would threaten the authority structure that grants them their position.
This is not a moral claim about participants. It is a structural claim about how authority structures with stake in kernel preservation generate predictable blindness. The constitutional originalist cannot easily acknowledge that their preferred interpretation is itself an evolution of original meaning. The church hierarchy cannot easily acknowledge that doctrinal development is doctrinal change. The institutional reformer cannot easily acknowledge that their proposed reform serves institutional interests. The identity-anchored person cannot easily acknowledge that their commitments have shifted operationally.
The disagreement between positions is data, not problem. Different perspectives on the same system produce unreconcilable accounts because the structural mechanisms they pass through differ across positions. The framework formalizes this through presheaf machinery developed in companion work — for readers interested in the mathematics, the framework treats observer positions as forming a category whose morphisms encode structural transitions, with classification data forming a presheaf that often fails the sheaf gluing axiom. The technical detail is not necessary to use the framework. The structural insight is that disagreement between positions can be a measurable feature of the system rather than evidence that someone is wrong.
The framework’s analytical position is also a position. The analyst sees patterns participants cannot access from their positions, but this exterior vantage is constitutive of the analytical move rather than incidental to it. The framework does not claim the analyst sees truth while participants see illusion. It claims the analyst’s position permits structural description that participant positions cannot produce, and that this is the same kind of positional difference the framework analyzes everywhere else.
How to use the framework
When examining any commitment system, four questions arise in order.
What is the kernel and the authority structure? Many disputes about commitment systems turn out to be disputes about which kernel is operative or which authority is legitimate. Specifying these is the precondition for further analysis.
Is drift occurring, and is it acknowledged? Drift is intrinsic. The question is whether the system can name it. The acknowledgment pattern is diagnostic of the configuration.
Which pattern is the system in, and is the pattern functional in this environment? The five patterns produce different failure modes. Environmental rate matters. Authority structure incentives matter. Functionality assessments are conditional on environment and incentive structure; they produce probability distributions over trajectories, not deterministic predictions.
Who has standing to acknowledge drift, and do they benefit from pretending it does not exist? This question is more prominent in relational and identity systems than in organizational systems because acknowledgment in those domains is structurally distributed. It generalizes to organizational systems through the structural distribution of who benefits from acknowledgment and who pays for it.
The framework provides scaffolding for thinking about these problems clearly. It does not resolve the substantive questions any particular commitment system faces. The US Constitution’s adaptation pathway is a political problem requiring political work. The Catholic Church’s relationship to doctrinal change is a theological problem requiring theological work. A specific marriage’s relationship to drift is a relational problem requiring relational work. The framework names what the questions are without claiming to settle them.
The framework as commitment system
The framework analyzes commitment systems. It is itself a commitment system, and treating it as one is part of the discipline the framework requires.
Its kernel is the structural claims developed in this and successor documents. Its authority structure consists of whoever adopts the framework and applies it. Its drift is real: the framework’s own commitments change as it gets applied, criticized, and revised. Its acknowledgment capacity is whatever revision discipline its practitioners maintain.
The framework’s own preference for marked revision over interpretive accretion is not a bias to be apologized for. The framework’s purpose is making drift visible. Marked revision is what drift becoming visible consists of. A 50-pound weight gain accumulates through increments below the threshold of recognition; the analytical move that makes the gain visible is necessarily a move that prefers marked recognition over absorbed change. The framework that did not prefer marked over absorbed would be a framework that could not make the distinction it exists to make.
The framework is not neutral about visibility. The framework is neutral about which patterns are viable. Interpretive accretion is adaptive across millennia in the right environments. Implicit practice is stable under appropriate conditions. Anchored fixity with interpretive accretion can produce extreme longevity. The framework’s analytical position is committed; the framework’s pattern assessments are not.
A demonstration: the environment-conditional stability of mathematics
The framework’s value test is whether it yields analytical moves that domain-specific framings don’t naturally produce. One demonstration.
Mathematics is conventionally treated as the paragon of stable knowledge — the canonical example of a tradition that has functioned for thousands of years without crisis. The framework’s prediction about marked-revision configurations supports this characterization. But the framework adds a calibration condition: stability is environment-relative. The same configuration that produces stability in one environment can produce stress in another.
Mathematics’ acknowledgment machinery is calibrated for a specific environment: proofs that can be checked by human readers within reasonable time, building on a literature small enough that experts can maintain working knowledge of their subfields. The environment is slow-changing — mathematical objects don’t change, foundational results stay valid, and amendment can proceed at the pace of human verification.
This environment has been changing. Computer-assisted proofs introduce results that no human can verify directly. The 1976 four-color theorem began the controversy; Hales’ 1998 proof of the Kepler conjecture made it acute when journal referees spent four years unable to verify the computer code, and the Annals of Mathematics ultimately published an abridged version while explicitly refusing to vouch for the computational part. The Coq formalization in 2005 settled the case but established that human verification was no longer sufficient for an emerging class of results. Subfields proliferate faster than experts can track them; Kevin Buzzard has argued in public talks that no single person understands the complete proof of Fermat’s Last Theorem, and that the mathematical literature may contain accepted results that are in fact incorrect — a claim he raises not as scandal but as motivation for formalization. Lean and other proof assistants are now being adopted not as supplements to human verification but as the only feasible way to verify some contemporary work; Peter Scholze’s perfectoid spaces work was verified through the Liquid Tensor Experiment.
The framework’s diagnosis: mathematics’ acknowledgment machinery is being stressed by an environment that now produces drift faster than human-paced verification can track. The proof assistant movement is bandwidth extension in response to environmental acceleration. The conventional framing — mathematics as paragon of stability — obscures what mathematicians are doing. The framework names it precisely: environmental acceleration is producing drift faster than the previous bandwidth could acknowledge, so the bandwidth is being extended to maintain the acknowledgment-rate match.
The result is intuitive once stated. Mathematicians extending their verification machinery in response to computational complexity are not abandoning the architecture; they are recalibrating it. But the conventional framing obscures what they are doing. The framework names it precisely.
The same analysis applies to any commitment system facing environmental acceleration. AI development is the contemporary case where the analysis matters most. AI training objectives, evaluation criteria, and safety specifications are kernels. Labs, deployment institutions, and regulatory bodies are authority structures. Distribution shift, reward hacking, goal misgeneralization, and emergent behaviors are drift. The framework predicts that AI development will face the same kind of bandwidth-extension challenge mathematics faces, complicated by the fact that AI development operates with capability acceleration on the operational side and rapid authority-structure evolution on the governance side simultaneously. Failure modes characteristic of this configuration include alignment targets becoming unstable mid-optimization, systems satisfying neither the framework they were trained against nor the framework they are deployed under, and operational meanings of “alignment” diverging across labs and regulators while each claims kernel fidelity. These predictions are concrete enough to be wrong, which is what makes them worth testing.
Why this matters
The standard debates about cultural change collapse into moralized disputes — tradition versus progress, fidelity versus innovation, original meaning versus living interpretation. These debates persist because they cannot be settled in their own terms. Whichever side wins one round, the other side returns with the same arguments restated.
The commitment systems framework reframes these debates as questions about architecture rather than values. The question is not whether drift is good or bad but whether the system’s authority structure can acknowledge the drift its domain actually produces, given the rate at which that domain changes and the incentives the authority structure has. Systems whose acknowledgment capacity matches their environment’s drift rate adapt without crisis. Systems whose acknowledgment capacity falls behind accumulate gap until something breaks.
This reframe has practical consequences across domains. It identifies which commitment systems are stable and which are unstable as structural rather than political questions. It distinguishes legitimate reform from rationalized continuity by asking whether reform changes the authority structure’s incentives or just relabels the kernel. It locates the source of recurring crises in architecture rather than personalities. And it provides a vocabulary for analyzing systems across domains that previously had no shared analytical framework.
The framework’s deepest point is simple: systems do not usually fail because people are bad or stupid. They fail because the mechanism for admitting change is broken or captured by people who benefit from things staying officially the same. Naming this mechanism is what the framework is for.
Appendix: Provisional Structural Refinements
The framework as stated above captures the core theory at the level of theory. The work of testing, operationalizing, and applying the framework will surface refinements that the theory paper cannot resolve from the armchair. The following provisional refinements emerged during the framework’s development and are documented here as input to subsequent operational work rather than as established theory. They are working notes, not theoretical commitments.
Drift decomposition. Drift as stated above bundles distinct dynamics that may benefit from decomposition along two axes. Source of drift: environmental (operational practice responds to environmental change), authority (the interpretive structure changes through personnel turnover or institutional evolution), or endogenous (the authority structure’s incentives produce reinterpretation independent of environmental pressure). Relationship to drift: unrecognized (perceptual or structural lag between drift and recognition) or exploited (participants who recognize the gap deliberately exploit it). Whether this decomposition produces orthogonal categories or remains analytically blurred depends on application work.
Three structural principles. Three statements about commitment systems may be operating as load-bearing principles rather than as implications of the patterns. Authority extends only as far as enforcement. An authority structure’s nominal power is bounded by its operational enforcement capacity; the gap between nominal and operational authority produces predictable failure modes. Drift is intrinsic. Every commitment system experiences drift because no kernel can fully anticipate the environment its operational practice will encounter. Acknowledgment that changes the system requires standing within the system. Outsiders can recognize drift; only insiders can ratify acknowledgment in ways that change what the system does. The principles are stated here as candidates rather than as established framework content.
Lifecycle phases. Commitment systems may be temporal entities passing through phases (formation, stabilization, operation, atrophy, renewal-or-dissolution) rather than static configurations. If so, the patterns above describe phase characteristics rather than timeless identities. The lifecycle dimension is invisible from inside the system — participants cannot know which phase they are in. The framework’s value would be shape recognition that aids probabilistic prediction, not deterministic outcomes. Phase identification from outside the system depends on observable indicators that operational work would need to specify.
Cross-cutting features. Several structural features may modify the base patterns. Decoupled formalization: a system’s formal kernel exists for purposes other than operational control (legibility, ceremonial completeness, external recognition) while operational authority sits elsewhere. Ritualized renewal: scheduled obligatory moments where the relationship between kernel and operational practice must be actively reconstituted (electoral cycles, documentary revision, physical reconstruction, sunset clauses); the central failure mode is performative renewal versus exercised renewal. Velocity mismatch: operational drift and authority drift operating at substantially different rates in either direction. Distribution of acknowledgment authority: concentrated (cathedral) versus distributed (bazaar) acknowledgment with different failure modes for each. Kernel encoding substrate: kernels can be encoded in ways admitting precise revision (formal text, mathematical notation) or in ways resisting precise operation (embodied knowledge, tacit practice, somatic response), with mismatches between acknowledgment mechanism and encoding substrate producing apparent acknowledgment without actual drift processing.
Default conditions. Two structural facts may obtain in most actual cases rather than being features systems have or don’t have. Nesting: almost every commitment system is nested in larger commitment systems and contains smaller commitment systems; the single-level system is the limit case rather than the typical case. Coupling: most commitment systems are coupled to other commitment systems whose dynamics affect their own; the environmental coupling the framework names is partly coupling to physical environment but mostly coupling to other commitment systems.
Observer parameters. The framework’s positional analysis may be parameterized through variables developed in companion work: power (what consequences a position can produce), exit options (what costs are involved in leaving the system), time horizons (what temporal patterns a position can observe), and spatial-temporal scope (what extent of the system a position sees).
Trajectory channels. Beyond standard absorption and rejection responses to drift, two additional channels may operate. Operational separation: inner systems continue without outer-system participation, producing parallel commitment systems with overlapping membership. Dormant-container activation: a system previously suspended but with preserved kernel and authority structure can be reactivated as alternative legitimacy when the active outer container loses standing. The Meiji Restoration’s reactivation of imperial legitimacy, France’s Estates-General reactivation in 1789, and identity deconstruction patterns where pre-conversion identity reactivates are candidate instances.
Cover story analysis. The structural prediction that positions with extraction stakes in kernel preservation generate predictable blindness may be testable when the predicted blindness is specific enough to be observed in advance. For a given configuration and position, the framework can in principle predict what drift the position should be unable to see and what cover story will replace it. The form would be: given X configuration plus Y features plus Z position, the predicted blindness is W with cover story V, and observation U would settle the question.
These refinements are provisional. The apparatus to test them does not yet exist. Building that apparatus is the next phase of work. What the refinements offer the theory paper is documentation of what subsequent operational work will need to address. What the theory paper offers the operational work is the structural foundation the refinements modify.
cafebedouin.org 