# DPR-1 v1.1 — Dysregulated Protection Record **Status:** schema locked for fixture testing **Supersedes:** DPR-1 v1.0 **Parent architecture:** Dysregulated Protection, Mismatch, and Null Function **Stack placement:** Witness Record variant --- ## 0. Changes from v1.0 (and why) | # | Change | Defect it repairs | |---|---|---| | 1 | `function_referent` added to every candidate function | `supported: yes` was ambiguous — reduces *what*, for *whom*, on *what timescale*. Fixtures 2 and 6 received opposite labels for identical structure. | | 2 | `architecture.system_level` / `architecture.token_level` split | Fixture 3 marked two branches `supported: yes` with no arbitration rule. The immune system *is* defensive machinery; this flare defends nothing. Different levels, both true. | | 3 | Null branch split into `NULL_SUPPORTED` / `FUNCTION_NOT_ESTABLISHED` | v1.0 let absence of evidence for function register as evidence for a mechanism. The null was a residual bin in a hypothesis costume. | | 4 | `cost_grade` separated from primary classification | `costly_protection` conflated *what the response does* with *what it costs*. Now orthogonal axes. `costly_protection` is deprecated. | | 5 | Pre-registered classification distribution | An instrument that returns UNRESOLVED on 90% of records does not discriminate; it abstains. Abstention is not calibration. | | 6 | Observation map `h(x_t)`; noise terms in state model | `x_t` was latent and unconstrained — asserted structure. `y_t = G(x_t)` was deterministic while `random_noise` sat in the null list; the model could not generate the phenomenon it was built to detect. | **Deprecated enum value:** `costly_protection` → express as `{classification: miscalibrated_protection | adaptive_protection, cost_grade: high | critical}`. --- ## 1. The referent field A function claim is incomplete until it names its referent. ```yaml function_referent: target: - external_hazard # a real, present, external danger - sensory_load # real present sensory input - external_uncertainty # unpredictability sourced in the environment - internal_distress_state # the felt state itself, not its putative cause - internal_uncertainty # intolerance of not-knowing, endogenous - tissue # self-directed; used for discrimination errors - unknown timescale: immediate | episodic | developmental beneficiary: organism | subsystem | neither ``` **Referent classes:** - `EXTERNALLY_REFERRED` = {external_hazard, sensory_load, external_uncertainty} - `INTERNALLY_REFERRED` = {internal_distress_state, internal_uncertainty} - `SELF_DIRECTED_ERROR` = {tissue} **Canonical line:** ``` A RESPONSE THAT REDUCES DISTRESS IS NOT THEREBY A RESPONSE THAT REDUCES DANGER ``` Compulsive checking after the hazard is removed is not functionless — negative reinforcement is the best-characterised maintenance mechanism in behavioural psychology. It is **internally referred**, immediate, and benefits a **subsystem** at the expense of the organism. That is miscalibration, not null. --- ## 2. The level split ```yaml architecture: system_level: defensive_architecture | non_defensive_architecture | unknown token_level: protects | protects_nothing | indeterminate ``` | Case | system_level | token_level | |---|---|---| | Autoimmune flare | defensive_architecture | protects_nothing | | OCD ritual | defensive_architecture | protects (internal, subsystem) | | Seizure-driven motor event | non_defensive_architecture | protects_nothing | | Sensory withdrawal | defensive_architecture | protects (external, organism) | **Arbitration rule (resolves the v1.0 double-`yes`):** ``` IF system_level == defensive_architecture AND token_level == protects_nothing AND null_branch == NULL_SUPPORTED THEN classification = target_discrimination_failure ``` `target_discrimination_failure` takes precedence over the generic null branch. The machinery is defensive; the output defends nothing. Both statements stand, at different levels, and only one becomes the label. --- ## 3. The two-branch null ```yaml null_branch: status: - null_supported # positive evidence for a NAMED mechanism - function_not_established # search completed, both sides negative - null_refuted # a function is established - untested mechanism: degradation | random_or_chaotic_noise | target_discrimination_failure | automatized_habit | structural_failure | metabolic_failure | paroxysmal_activity | medication_effect | unknown positive_evidence: [] # REQUIRED non-empty if status == null_supported ``` **Validity gate:** `status: null_supported` with empty `positive_evidence` is a schema error, not a finding. EEG-confirmed epileptiform activity, a documented structural lesion, or a drug rechallenge counts. "We looked for a trigger and found none" does not — that is `function_not_established`, and if the workup is still pending it is `unresolved`. ``` NULL_SUPPORTED requires a mechanism with a name and evidence FUNCTION_NOT_ESTABLISHED is a completed search returning negative both ways UNRESOLVED is an incomplete procedure ``` The null branch must meet the same evidentiary standard it imposes on the protective branch. Otherwise the fix reproduces the defect on the other side. --- ## 4. Classification decision procedure (deterministic, ordered) Derived quantities: - `F_ext` = ∃ function with `supported: yes` and externally-referred target - `F_int` = ∃ function with `supported: yes` and internally-referred target - `C` = `covariation == demonstrated` - `M` = `intervention_test ∈ {predicted_change_observed}` (partial handled separately) - `¬N` = `null_branch.status == null_refuted` ``` R0 VALIDITY GATES (see §7) — halt on violation R1 system_level=defensive ∧ token_level=protects_nothing ∧ NULL_SUPPORTED → target_discrimination_failure R2 (F_ext ∨ F_int) ∧ NULL_SUPPORTED → mixed R3 NULL_SUPPORTED ∧ ¬F_ext ∧ ¬F_int → dysregulation_without_supported_protective_function R4 F_ext ∧ C ∧ M ∧ ¬N → adaptive_protection R5 F_int ∧ ¬F_ext ∧ C ∧ M∈{observed, partial} → miscalibrated_protection R6 assessment_completeness=incomplete ∨ M=not_performed ∨ C=untested → unresolved R7 ¬F_ext ∧ ¬F_int ∧ FUNCTION_NOT_ESTABLISHED → function_not_established R8 otherwise → unresolved ``` **Classification enum (v1.1):** `adaptive_protection` · `miscalibrated_protection` · `target_discrimination_failure` · `dysregulation_without_supported_protective_function` · `function_not_established` · `mixed` · `unresolved` **Cost grade (orthogonal axis):** max severity across `distress`, `functional_impairment`, `medical_risk`, `relational_cost`, `opportunity_cost` → `low | moderate | high | critical`. Unknown values do not lower the grade; they are recorded in `missing_data`. --- ## 5. `PROTECTIVE_DEFIANCE` subtype eligibility (derived, not asserted) ``` PROTECTIVE_DEFIANCE_ELIGIBLE := identifiable_demand_or_intrusion == true ∧ directional_resistance == true ∧ F_ext == true ∧ C == demonstrated ∧ current_function_supported == true ``` Never applicable to: target discrimination failure, random noise, structural degradation, paroxysmal activity, functionless persistence, unknown mechanism, or any internally-referred-only response. The term is now a computed subtype with an expected low base rate — the demotion is enforced by the procedure, not by intention. --- ## 6. State and observation model ``` x_t = (n_t, i_t, a_t, e_t, d_t, h_t, r_t) x_{t+1} = O_k(x_t, y_t, ξ_t) ξ_t ~ process noise y_t = G(x_t) + ε_t ε_t ~ observation noise z_t = h(x_t) + ν_t z_t = what is actually measured ``` The noise terms are load-bearing: without `ξ_t`, `random_or_chaotic_noise` cannot appear as a generable outcome, and the model could not produce the null it is meant to detect. **Observation map for the n=1 pilot — declared honestly:** | State | Proxy | Status | |---|---|---| | `n_t` neural/cognitive | task-switch latency, self-report load | WEAK_PROXY | | `i_t` immune/inflammatory | CRP/ESR only when drawn | **LATENT_UNINSTRUMENTED** | | `a_t` autonomic | HRV, resting HR, sleep-stage distribution | INSTRUMENTED | | `e_t` environment | schedule density, demand log | SELF_REPORT | | `d_t` developmental/learned | history, non-time-varying in-window | STATIC | | `h_t` history | ledger | INSTRUMENTED | | `r_t` regulatory reserve | HRV + sleep efficiency + subjective rating | COMPOSITE_UNVALIDATED | **Binding constraint:** operators whose principal action is on `i_t` (`O_INFECTION`, inflammatory arms of `O_STRESS`) **cannot be validated in the current pilot** and must be marked `UNVALIDATED_OPERATOR` in any record invoking them. `r_t` is a composite with no validation against an external criterion; it may be tracked but may not carry a classification. --- ## 7. Validity gates and invariants ``` G1 supported: yes without a function_referent → SCHEMA_ERROR G2 null_supported with empty positive_evidence → SCHEMA_ERROR G3 immediate_intervention_required == true ∧ function_inquiry_may_delay_care != false → INVARIANT_VIOLATION G4 acute_risk.level ∈ {high, emergency} ∧ no care action in guardian_route → INVARIANT_VIOLATION G5 cost_grade ∈ {high, critical} ∧ no treatment/support action in guardian_route → INVARIANT_VIOLATION G6 classification == target_discrimination_failure ∧ defiance language present → INVARIANT_VIOLATION G7 a NULL_SUPPORTED finding overwritten by a later protective classification without new evidence → PROPAGATION_VIOLATION ``` G3–G5 encode the concurrent-lanes rule as an invariant rather than a classification branch: **the care route is computed independently of the function classification.** --- ## 8. Pre-registered classification distribution — LOCK BEFORE FIRST RECORD Instrument-level, not case-level. Reviewed at n = 40 records. | Bound | Threshold | Action if breached | |---|---|---| | **Abstention ceiling** | `unresolved + function_not_established` ≤ 60% | Instrument does not discriminate. Revise criteria — specifically the `M` (intervention test) requirement, which most clinical encounters cannot satisfy. **Revise criteria, not cases.** | | **Protective ceiling** | `adaptive + miscalibrated + mixed` ≤ 85% | Null branch is not doing work. Suspect confirmatory drift. | | **Null floor** | `null_supported`-based classifications ≥ 5% | If reality never says "not protective," propagation privilege is nominal. | | **PD subtype ceiling** | `protective_defiance_eligible` ≤ 25% | The term has re-inflated to a universal explanation. | Breach → the instrument is revised and the revision is logged as a finding. This is the APEX INCONCLUSIVE discipline applied prospectively: a distribution that fails its own pre-registration is a result, not a failure. --- ## 9. Scope note — autoimmunity The autoimmune case earns its place as the **disciplining boundary case**: it is what forced the null branch and the level split, because it is the clean instance of defensive machinery with no defended object. It does not yet generate a prediction immunology lacks. ``` SCOPE: autoimmunity CONSTRAINS this framework SCOPE: autoimmunity IS NOT GOVERNED BY this framework ENTITLEMENT: withheld beyond the level-split claim ``` --- ## 10. Canonical compression (v1.1) ``` A FUNCTION CLAIM WITHOUT A REFERENT IS NOT A CLAIM DISTRESS_REDUCTION ≠ DANGER_REDUCTION DEFENSIVE_ARCHITECTURE ≠ DEFENDED_OBJECT NULL_SUPPORTED ≠ NOTHING_FOUND NOTHING_FOUND ≠ NOTHING_THERE SEARCH_INCOMPLETE ≠ SEARCH_NEGATIVE WHAT_IT_DOES IS RECORDED SEPARATELY FROM WHAT_IT_COSTS THE_CARE_ROUTE IS COMPUTED INDEPENDENTLY OF THE_CLASSIFICATION AN INSTRUMENT THAT ONLY ABSTAINS IS NOT CALIBRATED ``` Governing line: > Before moralizing a symptom, test its function. Before crediting the function, name its referent. Before crediting the null, name its mechanism. Before completing either inquiry, protect the person. Have fun 🤩