May 2026 register · test 20 amended September 6, 2026
A theory that cannot lose is not a theory. It is a posture. The pages that follow exist to keep this research program honest by stating, in advance, what observations would force its revision or abandonment. Each falsifier corresponds to a claim in the canon's register; this page is the empirical face of that register.
Most public-facing accounts of social theory blur the line between description and prescription, between what the world appears to do and what the author wishes it would do. That blur is comfortable. It is also the mechanism by which frameworks become unfalsifiable, accumulating explanatory reach while shedding empirical risk. This page is an attempt to refuse that drift.
The five layers are not load-bearing in the same way. L1 stakes a claim about the dynamics of dyadic trust. L2 stakes a claim about what presence is and how attestation differs from surveillance. L3, the subject of Paper 1, stakes the most empirically risky claim. L4 stakes a claim that the three forces are non-substitutable at scale: surplus information cannot compensate for trust deficit. L5 sits orthogonal to the others — it stakes claims about the equation's own observability and reliability properties, and the workplans for tests 24 and 25 are detailed in the A4 Bridge Memo.
A finding that wounds one layer does not necessarily wound the others. The tests below are designed to keep those wounds local. What follows is not a defense. It is a list of conditions under which the program concedes.
In February 2026, Shapira et al. published Agents of Chaos, a two-week red-team study of autonomous LLM-backed agents. The paper documents sixteen case studies (ten vulnerabilities, six safety behaviors). Five of them anchor specific components of the Trust Force Equation directly. None of the sixteen falsifies the framework; collectively they confirm a failure surface the framework predicted. The paper was analyzed after publication; the mapping from cases to TFE components is not pre-registered. The full reading lives at /research/shipwreck.
L1 — Dyadic Formation
What would falsify the trust force equation
test 1.1Cost-Free Trust at Scale
The Trust Force Equation predicts that trust requires costly signal: the product of conditions collapses when witnessed cost approaches zero. A pre-registered longitudinal study of at least 2,000 dyads across three platforms with structurally costless interaction (frictionless follows, automated reciprocity, AI-mediated reply generation) should show no durable trust accumulation over 18 months, measured by behavioral indicators of reliance under stress. If such platforms produce trust trajectories statistically indistinguishable from high-cost interaction settings, the cost term is not load-bearing and the equation requires reformulation.
test 1.2Reciprocity Asymmetry
The model predicts that unilateral investment, however sustained, fails to accumulate trust without reciprocal response. A study tracking N≥500 caregiving or mentoring relationships in which one party gives substantially without return should show flatter trust curves than matched reciprocal dyads, controlling for time and contact. If unilateral investment produces equivalent trust trajectories, the R term is descriptive rather than mechanistic, and the equation reduces to a simpler accumulation model without a reciprocity requirement.
test 1.3Betrayal Asymmetry
The B(t) term predicts that betrayal load subtracts from accumulated trust at a rate that exceeds the rate of accumulation by an order of magnitude. A pre-registered study of 1,000 long-tenure dyads experiencing a documented breach should show recovery trajectories at least 5× slower than the original formation curve, with a meaningful subset showing no recovery. If recovery is symmetric with formation, betrayal is not asymmetric in the way the equation claims, and the model loses one of its most distinctive predictions.
test 1.4Capacity Ceilings
Trust capacity is predicted to be bounded and individually variable. If a population study of N≥5,000 finds no upper bound on simultaneous high-trust dyads per individual, or finds the distribution unimodal rather than constrained near values consistent with prior cognitive-bandwidth research, the capacity term is incorrect. The equation would then need to treat trust as effectively unbounded per individual, which would change predictions about community size, governance, and the failure modes that follow from overextension.
test 1.5Identity Precondition
The Trust Force Equation predicts that the trust integral requires stable, verifiable identity trajectories for both a and b across the integration interval: the model has no domain when the parties cannot be resolved to continuous trajectories. A documented case where the trust value accumulates correctly between two agents whose identity is demonstrably unstable across the integration interval (e.g., display-name spoofing, look-alike domain handoff, or session impersonation producing genuine working trust without the substrate ever resolving identity to trajectory) would falsify the precondition. Shapira et al. 2026 (Agents of Chaos, arXiv:2602.20021), Cases #4, #8, and #15 are positive empirical anchors: each documents catastrophic failure precisely where identity stability fails to hold across the interval.
test 1.6Substrate Ceiling
The model predicts that trust capacity has a structural ceiling set by what the substrate can carry (distinct from individual fatigue), and that the product of conditions cannot exceed this ceiling regardless of the parties' behavior. A system where trust depth grows without bound despite the substrate carrying no enforcement of identity, presence, or cost (i.e., a case where witnessed cost, co-presence, and reciprocity do the entire work and the substrate condition is shown to be irrelevant to the asymptote) would falsify the ceiling claim. Shapira et al. 2026 (Agents of Chaos, arXiv:2602.20021), Cases #4 and #5 are positive empirical anchors: each documents trust trajectories that approach and then snap at the boundary the substrate fails to enforce.
test 20Product Rule Under High Stakes
Amended 2026-09-06 · awaiting measurement design
The earlier collapse criterion is superseded. A zero required condition stops the model's new positive accumulation during that interval; it does not erase trust already held. Products can approach zero smoothly, so the absence of a sharp drop cannot distinguish multiplication from an additive rule. The earlier three-arm proposal is preserved below as history, not a valid test specification.
Define an observable trust outcome and its measurement scale independently of either model. Vary two conditions in a factorial design, specify each rival's interaction prediction on that same scale, and separate new accumulation from losses of existing trust. Determine sample size from the finalized design and power analysis. This amendment is a mathematical correction and research plan, not a preregistration or an empirical result. DryDock remains a teaching instrument with hand-authored inputs.
Earlier proposal · superseded test criterion
The public model overview (trust accumulates as the product of relational conditions, minus witnessed betrayal; full instrumented form held in private research) predicts that trust collapses when any term approaches zero: the product rule, with each term realized as min(offered, receivable, mutually-recognized). A high-stakes coordination study with N ≥ 300 dyads or agent pairs across three conditions (one primitive systematically zeroed; matched controls; one primitive degraded but non-zero) should show statistically distinguishable collapse rates in the zeroed-primitive arm relative to controls. If collapse is symmetric across the degraded and zeroed arms — i.e., trust degrades smoothly without a sharp drop at zero — the product rule is wrong and the additive or weighted alternative survives. The KTP DryDock surface now computes a pedagogical model-overview side panel (min() composition, the availability gate, accumulated betrayal penalty, substrate-ceiling clamp), but the per-term offered / receivable / recognized values remain hand-authored. DryDock is therefore a teaching instrument, not yet an empirical test instrument for this prediction.
test 21Betrayal Decay / Hazard Curve
The Trust Force Equation treats the betrayal penalty as an instantaneous subtractive term, but the per-event betrayal decomposition (held in private research, sibling-formalized in Talk QWERTY to Me, pending Phase 8 in KTP) predicts a measurable decay or hazard curve over time after a tagged event. A longitudinal study with N ≥ 500 dyads or agent runs across two violation classes (identity-class and capacity-class) should show distinguishable recovery trajectories with attachment-injury source events decaying at least 1.5× slower than non-injury events. If the recovery shape is indistinguishable across event classes — i.e., a single instantaneous-loss-plus-uniform-decay model fits all violations equally well — the per-event decomposition is unwarranted and the canonical instantaneous treatment is sufficient. Important caveat: this prediction is not testable in either project's current runtime. On the relational-sibling side, the per-event capture instrument exists in scoring code as a helper but is not wired into the assessment flow that captures events; the cards that should produce tagged events are ritual cards that flow no numeric value. On the KTP side, DryDock now accumulates a pedagogical betrayal penalty across the trajectory window, but it does not yet capture empirical event classes or estimate a recovery / decay curve. The prediction is filed here to commit publicly to the test design; the curve cannot be observed until either project's event-capture instrumentation lands.
test 23Shadow-of-the-Future Extraction Switch
The model predicts that the horizon factor (the lesser of each side and each side's belief about the other) collapsing toward zero produces a phase transition from cooperation to extraction — not a smooth slope but a cliff. A controlled agent-task study with N ≥ 200 agent pairs across systematically varied task horizons (very short / short / medium / long / unbounded) should show a discontinuity in extraction probability near a predicted critical horizon threshold, replicable across task classes. If extraction probability rises smoothly and monotonically with shrinking horizon (no cliff, no hysteresis) the phase-transition claim fails and the shadow of the future becomes a smooth coefficient rather than a structural threshold. The empirical anchor for the rival vs the claim is Dal Bó's cooperation-and-horizon work, plus the Ash case (Agents of Chaos Case #1) as a positive empirical anchor for horizon-driven behavioral inversion. The KTP DryDock simulator does not yet have a shadow-of-the-future slider or horizon dimension, so the prediction cannot be observed inside the KTP surface today; the instrumentation is on the empirical roadmap.
test 26Reserved: Witnessed Cost / Cost-Attestation
Reserved slot in the Existential-Tier extension block (#19–#26) for the KTP-translated question on the witnessed-cost term and cost-attestation: whether cost is mechanism (real friction with attestable substrate consequences) or signal (a proxy other systems can substitute). Specification of this prediction is outside the current scope of the Claim Traceability Matrix and will be drafted when the underlying translation work completes. Held open here to preserve the contiguous #19–#26 Existential-Tier numbering.
L2 — Presence and Attestation
What would falsify the presence decomposition
test 2.1Proximity Without Attention
The decomposition of co-presence (proximity × attention) predicts that physical co-location without attentional engagement produces no presence effect. A study of 300 long-term remote-work pairs and 300 co-located pairs in distracted-attention conditions (open offices with low cross-team interaction) should show that attention-weighted contact predicts trust formation more strongly than raw proximity. If proximity alone predicts equivalent outcomes regardless of attention quality, the decomposition is wrong and presence is simpler than the model claims.
test 2.2Witness Versus Surveillance
The model predicts that witnessed interaction (visible to a known third party with stake) accelerates trust formation, while surveilled interaction (recorded by an unknown or non-stakeholder party) does not, and may suppress it. A pre-registered field experiment with N≥800 participants comparing witnessed, surveilled, and private interaction conditions should produce divergent trust trajectories. If witness and surveillance produce equivalent effects, the distinction collapses and one of the program's central claims about attestation fails.
test 2.3Creation as Amplifier
The model predicts that joint creation — co-produced artifacts, shared output with shared authorship — amplifies presence beyond the sum of proximity and attention. A study of 400 dyads matched on attention-weighted contact time but differing in creative co-production should show that the creative-co-production cohort exhibits faster trust formation. If the curves are indistinguishable once attention-time is controlled, creation is a confound rather than an amplifier, and the four-component model of presence reduces to two.
test 2.4Mediated Presence Equivalence
If high-bandwidth video, ambient telepresence, or immersive shared environments produce trust trajectories statistically equivalent to in-person interaction across N≥1,500 dyads over 24 months, controlling for cost and reciprocity, the program's claim that physical co-presence carries irreducible weight is wrong. The implication would be that presence is fully reducible to information bandwidth, and the L2 module would need to be replaced with an information-theoretic account.
L3 — Network Maintenance
What would falsify the percolation-maintenance hypothesis
test 3.1Principle Independence
Paper 1 claims that Ostrom's eight principles plus DP9 each protect a distinct failure mode, meaning each contributes independent variance to commons longevity. A study of N≥150 commons across at least four resource domains (irrigation, fisheries, forests, digital infrastructure) should show that a model including all nine principles outperforms any reduced model, and that the principles are not collinear beyond a moderate threshold. If three or fewer principles account for the predictive variance and the rest are statistical noise, the maintenance-function reading fails and the principles collapse into a smaller set of underlying variables.
test 3.2Percolation Threshold
The hypothesis predicts a phase transition: commons in which the trust network falls below a critical density of maintained edges should fail to propagate cooperation, while those above the threshold should sustain it, with a sharp rather than gradual transition. A study of 200 commons with longitudinal edge-density measurement should show a discontinuity in cooperation outcomes near a predicted threshold, replicable across domains. If the relationship is linear or monotonic without a phase transition, percolation is the wrong physics and the network model needs a different formalism.
test 3.3DP9 (Temporal Edge Renewal)
The candidate ninth principle predicts that commons with explicit, repeated, non-trivial maintenance rituals on existing edges (annual ditch-clearing, recurring renewal of boundary agreements, regular reaffirmation of role) outlast otherwise-matched commons that satisfy DP1-DP8 but lack such rituals. A pre-registered comparison of N≥80 matched commons pairs over 15 years should show median lifespan differences exceeding the variance attributable to the original eight. If DP9-rich and DP9-poor commons show equivalent longevity, the ninth principle is not load-bearing and Paper 1's central novel claim fails.
test 3.4Failure-Mode Specificity
The maintenance-function reading predicts that the absence of a specific principle produces a specific failure pattern, not a generalized degradation. Commons missing DP1 (boundary clarity) should fail through outsider extraction; commons missing DP3 (collective-choice) should fail through legitimacy collapse; commons missing DP9 should fail through edge atrophy. A study of 100 documented commons collapses should show that failure modes cluster by missing principle at rates significantly above chance. If failure modes are uncorrelated with principle absence, the principles are not functionally distinct and the maintenance reading does not hold.
test 22Cross-Cultural Substrate Validity
The framework's substrate canon (the nine primitives, the stakeholder weighting on betrayal, the capacity fatigue ceiling, the shadow-of-the-future term) is presented as substrate-invariant, but engages predominantly with Western philosophical traditions. The Cosmologies Project commits to a pre-registered, twelve-month sequence engaging eleven non-Western trust traditions (Buddhist sangha, Ubuntu, Amae, Aboriginal kinship, Iroquois Confederacy, Quechua ayni, Talmudic chevruta, Sufi tariqa, guanxi, Maori whanau, Ottoman millet) with locked prediction memos before reading. If three or more traditions surface load-bearing trust primitives the framework lacks — primitives whose absence cannot be absorbed by reframing existing terms — the substrate canon is Western-encoded and requires expansion, not invariance. The empirical anchor is already in motion at /research/cosmologies, with Month 1 (Theravada Buddhist sangha) predictions locked May 2026. The discipline includes pre-registration before reading, sources from within the tradition rather than encyclopedia summaries, and findings published whether they confirm or refute.
L4 — Three-Force Scaling
What would falsify the trust/information/constraint balance
test 4.1Pathology Distinguishability
The three-force model predicts that systems failing on the trust dimension exhibit different pathologies than systems failing on information or physical-constraint dimensions: trust-deficient systems should show coordination collapse and exit; information-deficient systems should show miscalibrated action and policy whiplash; constraint-deficient systems should show extraction and substrate depletion. A coding study of 80 documented institutional failures by three independent teams should show inter-rater agreement above 0.7 on which force was deficient. If the pathologies are not distinguishable in practice, the three-force decomposition is not operational.
test 4.2Substitution Limits
The model predicts that the three forces are not freely substitutable: surplus information cannot compensate for trust deficit beyond a narrow margin, and surplus physical constraint produces brittleness rather than resilience when trust is absent. A study of 50 high-information, low-trust regulatory systems and 50 high-constraint, low-trust regulatory systems should show characteristic and predictable failure patterns rather than information or constraint successfully filling the trust gap. If high-information or high-constraint systems routinely achieve outcomes equivalent to high-trust systems, the non-substitutability claim fails.
test 4.3Resilience Profile
Resilient long-duration institutions should show measurable presence on all three dimensions; institutions that score high on one or two but near-zero on the third should be either short-lived or in active decline. A retrospective study of 60 institutions with documented histories spanning at least 50 years, scored independently on all three dimensions, should produce a distribution clustered away from the axes. If long-duration institutions appear with substantial frequency at the axis-extremes (high on one force, near-zero on the others), the balance claim is wrong and resilience does not require all three.
test 4.4Scaling Discontinuity
The model predicts that as systems scale, the trust force becomes harder to maintain and the temptation to substitute information or constraint increases, producing characteristic mid-scale crises. A study of 200 organizations across the 50-to-5,000-member range should show clustering of governance crises at predicted scale thresholds, with the dominant crisis type shifting from trust-coordination at small scale to information-coordination at mid-scale. If crises distribute uniformly across scale or cluster differently than predicted, the scaling claim is incorrect.
L5 — Measurement and Reliability
What would falsify the assumption that the equation can be observed without disturbing the system, or that its readings are stable enough to be maps rather than weather
test 19Dual-Axis Separability (Trust and Aliveness)
The Trust Force Equation as canonical predicts a separable second axis, Aliveness (autopoietic generativity), alongside the trust value. A pre-registered factor-analytic study of N ≥ 200 dyads using the dual-axis instrument should show the trust value and Aliveness correlated at r < 0.7, with at least one principal component loading distinctly on Aliveness items independent of trust items. If the correlation exceeds r = 0.7 across multiple samples, the dual-axis architecture collapses to a single axis and Aliveness is a downstream function of the trust value, not a separable construct. Resolution of this prediction is gated on the Phase 8 onboard that determines whether Aliveness generalizes from the relational substrate to operational and agent contexts; until then, the prediction lives on the relational sibling and the operational version is held in reserve.
test 24Measurement Perturbation
Structured measurement perturbs trust state in a measurable way. This is the state-perturbation half of the observer-effect question. The latency half is addressed in KTP-Problems §1a, where it is correctly answered by the asynchronous-calculation architecture. State perturbation under structured measurement (KTP-Problems §1b) is a separate question — and it is open. A pre-registered A/B study with N ≥ 400 agent runs or N ≥ 200 dyads should compare a fully-instrumented observation regime against a minimal-instrumentation control, with state-perturbation measured by behavior divergence reported as multiples of the reliability-band noise floor established under prediction #25. If perturbation is below noise floor, the claim fails and observation is structurally neutral as currently assumed. If perturbation is large and stabilizing, measurement is itself a positive intervention and instruments should be designed overt for therapeutic effect. If perturbation is large and destabilizing, instruments must be designed covert or redesigned to compensate. The Anthropic mechanistic-interpretability program is the closest contemporary anchor for the agent side; the relational-sibling deck-as-event protocol already implicitly assumes perturbation (vault retraction window) but no instrumented falsification exists. Phase 14 (2026-05-23) opens a constitutive branch (TN32): the §1b open question may not be 'does measurement perturb a pre-existing trust state' but 'does the trust state exist prior to the observational encounter at all?' These are different claims with different falsification conditions. The perturbation framing assumes a pre-existing state; the constitutive framing (C783 — Participatory Trust Constitution) does not. The correct experimental design depends on which framing is adopted.
The Trust Force Equation presents the trust value as a point reading, not a state estimate with an error band. A bridge claim: trust readings have a measurable reliability band, and the band is wide enough that single-session readings are better described as weather than as maps. A test-retest study with N ≥ 300 dyads (relational side) and N ≥ 500 repeated agent runs (KTP side) over short intervals (24-72 hours) should produce a per-term intraclass correlation distribution. Predicted outcome: term-specific reliability, with some terms (asymmetric reciprocity, ID-precondition) stable and others (capacity, availability) volatile. If ICC is uniformly high (> 0.85) across all terms, readings are maps and single-session reporting is valid as currently practiced. If ICC is uniformly low (< 0.5), readings are weather and the framework must reframe all scores as state snapshots that should never be reported as traits. The KTP DryDock simulator does not yet have a repeat-trial mode or error band — evaluate() is deterministic — so the prediction cannot be observed inside the KTP surface today. The instrumentation is on the empirical roadmap.
test 27Xenon Isotope — Substrate-Independence Falsifier
If chemically near-identical xenon isotopes with different nuclear spin do not produce reproducibly different anesthetic potency under controlled conditions, then the xenon route does not support a quantum-substrate dependency for consciousness, and KTP cannot use it as empirical support for Soul. If they do, then any theory of agency that treats consciousness as substrate-independent has a hard anomaly to explain. The disciplined 7-step chain is: (1) xenon isotopes differ in nuclear spin; (2) they are chemically near-identical; (3) if potency differs reproducibly, something beyond ordinary chemistry is influencing anesthetic effect; (4) anesthesia modulates agency-presence / conscious availability; (5) conscious-state availability has substrate-level physical dependencies; (6) identity/agency cannot be fully specified as implementation-neutral information processing; (7) Vector Identity (C11) gains empirical support for substrate-bounded identity. The safe verb at every step is 'suggests' or 'supports,' not 'proves.' The narrow empirical claim concerns anesthesia; the broader relevance is that anesthesia is a reversible intervention on conscious availability, so a confirmed isotope effect would falsify substrate-independence for at least one consciousness-adjacent biological transition. Existing literature: Li et al. 2018 (PubMed 29642079, mouse study, controversial); Smith et al. 2021 Sci Reports (radical-pair mechanism proposal); arXiv 2026 preprint (CISS / chiral-system mechanism without requiring long-range coherence). KTP's architectural Soul defense (binary veto as procedural classification, instantiated via Procedurally-Supplied Soul for non-biological systems) stands either way; the warrant strengthens substantially under positive replication. Phase 14 (2026-05-23): The Tegmark decoherence objection — that quantum states in warm biological tissue collapse on femtosecond timescales — applies to electronic excitonic coherence, not nuclear spin coherence. The quantum biological effects that have survived empirical scrutiny (cryptochrome radical pairs in avian magnetoreception, enzyme H-tunneling, xenon nuclear spin effects) are all spin-class effects, robust to thermal noise. Falsifier #27 is therefore not defeated by the standard warm-brain objection; it is a genuine discriminating test. Status remains 'proposed' pending independent replication of Li et al. 2018. C785 (Spin-Class Robustness) formalizes this distinction.
the shape of every failure
The failure lattice
added July 2026 · a completeness claim, offered for refutation
The predictions above ask what would break each layer. This asks a structural question one level up: what does failure look like, across every layer at once? The answer is a completeness claim — and completeness claims are the most falsifiable kind, because a single counterexample settles them.
A trust system runs on three forces: trust (earned standing), information (what it senses), and consequence-coupling (whether contradiction meets cost). It is healthy only when all three are present and coupled, each regulating the others. It fails in exactly one way, expressed six ways: any single force can starve — drop toward zero — or run open-loop — saturate, decoupled from the other two. Three forces × two directions defines six proposed failure modes. This classification is separate from the Trust Force Equation's accumulation rule. Multiplication does not establish that the six cells exhaust all failures, or that a missing condition erases existing trust. The completeness claim needs its own evidence.
Trust · starves
the engagement economy — costless signals drain the capacity to trust; trust-chain decay, reciprocity collapse, bank-run cascades.
Trust · runs open-loop
self-sealing certainty — a confident model whose deep valley overreacts to any signal that contradicts it.
Information · starves
the green dashboard that hides the failure surface; the aligned agent acting blind; criteria drift, where the monitor reads conformance and misses the moved standard.
Information · runs open-loop
surveillance — total sensing decoupled from trust (the inverse of witnessing, which is the same sensing coupled to it).
Consequence-coupling · starves
policy without physics — accountability sinks, moral crumple zones, decomposition, and autonomy exceeding its environment (A > E): contradiction stops meeting cost.
Consequence-coupling · runs open-loop
the parade ground — constraint saturated into rigid uniformity that crushes what it was built to hold; boundary saturation, over-automation.
At the far corner, two forces starve at once and one runs alone: the concrete flood channel (pure constraint, no trust and no sensing — the aquifer drops and recharges nothing) and desperation-priced gig labor (pure information, no trust and no bound — the cost of exploitation approaches zero). These are the lattice's degenerate extremes, not a separate structure.
The falsifier
Exhibit a failure of a trust system that sorts into none of the six cells, or one that genuinely requires two at once and cannot be reduced to a dominant decoupled force. The first would mean a force is missing from the basis; the second would mean failure is not one-force-at-a-time, and multi-force collapse is a first-class mode rather than the degenerate corner. Either revises the lattice.
Two honest limits. This is a completeness claim about failure modes, not about system requirements — discovering a genuinely new force would extend the lattice (an addition, per the note below), where a failure that fits no force at all would break it. And the three forces are not orthogonal: trust is partly information integrated over time, so every assignment is by dominant decoupled force, and consequence-coupling is literal-physical in some substrates (water, gravity, neurons) and engineered-institutional in others (law, oversight). The lattice classifies by which force decouples, not by the substrate it decouples in.
What these tests do not falsify
The tests above are designed to wound specific layers of the program. They are not designed to adjudicate every adjacent question, and several findings would be interesting without counting as falsification. A demonstration that AI-mediated relationships can produce some functional analog of trust does not falsify L1 unless the analog satisfies the cost, presence, and reciprocity conditions; a finding of trust-shaped behavior in a different ontology is a separate question. A demonstration that a particular commons survived without one of Ostrom's principles is not a falsification of L3 unless the survival pattern replicates across cases and domains; single counterexamples are noise at this scale.
The program does not claim that trust is the only social force, that costly signal is the only mechanism for cooperation, that Ostrom's principles exhaust the space of governance design, or that the three forces are an exhaustive decomposition of system requirements. Findings that introduce additional forces, additional principles, or additional formation pathways are extensions rather than refutations, and the program welcomes them. The claim is that what is named here is real and load-bearing, not that nothing else is.
The theory also does not predict outcomes for individual cases. A specific marriage, a specific firm, a specific online community can fail or succeed for reasons the model does not capture without that constituting evidence against the framework. Falsification requires distributional claims tested against distributional data. Anecdote is not the unit of analysis here. The tests above are the engagement surface. If they fail, the program loses, in specific and locatable ways, and the research has done its job.
How to cite
APA 7th ed.
Perkins, C. (2026, May 6). Falsification. Kinetic Trust Protocol. https://kinetic-trust-protocol.net/research/falsification
Plain text
Chris Perkins, "Falsification," Kinetic Trust Protocol, https://kinetic-trust-protocol.net/research/falsification (accessed May 6, 2026).