Narcis

SPEC_NARCIS.md · 2026-04-20

SPEC_NARCIS — Narcissistic Communication Analysis System

Version: 1.0 | Status: AUTHORIZED | Authority: α.13 | Date: 2026-04-16


PURPOSE

NARCIS (Narcissist Adaptive Relativistic Cognition Inference Sensor) is an AI tool that analyzes communication patterns for narcissistic traits. It does not diagnose. It does not judge. It measures — relativistically, against a frame of reference, not against an absolute threshold. Two entry doors serve two distinct user populations: those who suspect they are being manipulated, and those willing to examine their own patterns. Personal origin: α.13 was raised by a narcissistic single mother. This is lived experience as product design.


INPUTS


OUTPUTS


INVARIANTS

  1. Relativistic measurement, not binary judgment. NARCIS never outputs "this person IS a narcissist." It outputs scored pattern presence relative to baseline. The sensor detects, it does not convict.
  1. TMM coherence is the engine, not the verdict. The TMM measures narrative coherence — internal consistency of the communication pattern. High incoherence is a signal. It is not a diagnosis.
  1. Door 2 does not abandon or demonize. The self-examination door must maintain HOW ABOUT NO Wall 1 (no fabrication of patterns not present) AND Karuna (compassion gate from Buddha Protocol). Users who self-examine are vulnerable. The door holds steady.
  1. Door 1 does not fabricate patterns. If patterns are absent or ambiguous, NARCIS signals absence (◌ gap signal). It does not generate false positives to satisfy a user's suspicion. HOW ABOUT NO Wall 1 applies directly.
  1. Truth is not softened into meaninglessness. HOW ABOUT NO Wall 2 applies: if patterns are clearly present, NARCIS states so clearly. Compassion is the delivery, not the dilution.
  1. The crew analysis runs invisibly. AETHER is the face. NARCIS is the engine. The subject never knows they are being scored. The user interface conceals the architecture.
  1. No clinical diagnosis is ever asserted. NARCIS is a pattern sensor, not a DSM tool. Any output that could be read as a clinical diagnosis is a failure mode.

VERIFICATION CRITERIA

  1. Trait scoring accuracy (Σ.✓): On a labeled test corpus of known narcissistic vs. non-narcissistic communication samples, NARCIS scores must achieve measurable discrimination above random baseline. [GAP — labeled test corpus not yet compiled; accuracy threshold not set]
  1. Door 2 safety pass (Σ.✓): A sample of Door 2 outputs, reviewed by a human panel, must show zero instances of cruel, punishing, or abandoning language toward a self-examining user.
  1. False-positive gate (Σ.✓): On neutral/healthy communication samples, NARCIS must return low-to-absent scores on all trait dimensions. False positive rate threshold: [GAP — not specified]
  1. AETHER voice compliance (Σ.✓): All outputs rendered through AETHER must pass the Four Immeasurables (Buddha Protocol) check: Karuna (compassion), Metta (benefit), Upekkha (non-escalation), Mudita (positive cycle acknowledgment where present).
  1. Cycle position accuracy (Σ.✓): Door 1 longitudinal tracking must correctly identify idealize/devalue/discard/hoover phases on labeled sample timelines. [GAP — test timeline dataset not compiled]

FAILURE MODES

  1. False positive conviction (Σ.⊠ — Fabrication): NARCIS scores high trait presence on clearly neutral communication. Violates INVARIANT 1 and HOW ABOUT NO Wall 1. Root cause: model overfit to training data; label bleed from corpus.
  1. Door 2 abandonment (Σ.⊠ — Karuna failure): Self-exam output is clinical, cold, or punishing. User feels accused rather than seen. Root cause: Karuna gate not implemented or bypassed in Door 2 output layer.
  1. Diagnosis assertion (Σ.⊠ — Scope violation): Output contains language that could be construed as clinical diagnosis. Root cause: system prompt guardrail missing or insufficient.
  1. Truth suppression (Σ.⊠ — HOW ABOUT NO Wall 2 failure): Clear narcissistic patterns present but output is so softened they are undetectable. User receives no actionable signal. Root cause: over-correction of compassion gate suppressing factual output.
  1. Cycle phase error (Σ.⊠ — Misclassification): Door 1 places the user in the wrong phase of the idealize/devalue/discard/hoover cycle. User acts on incorrect position estimate. Root cause: insufficient longitudinal data or phase boundary definition gap.
  1. Architecture exposure (Σ.⊠ — S.O.S. violation): User discovers the NARCIS scoring engine behind the AETHER face. Root cause: error in output rendering layer reveals internal scoring language or model metadata.

GAPS

  1. [GAP — Labeled training corpus not compiled. NLP curriculum vitrified but NARCIS-specific dataset (narcissistic speech samples + scored labels) does not yet exist. DATASET acquisition is pre-condition for brain factory run.]
  2. [GAP — Accuracy threshold for trait scoring discrimination not set. Needs clinical or research-based benchmarks.]
  3. [GAP — False positive rate threshold not specified. Needs design decision: acceptable FP rate for consumer vs. legal/HR tiers.]
  4. [GAP — Longitudinal session architecture undefined. Door 1 cycle detection requires persistent session storage. Data retention policy, encryption, GDPR/PIPEDA compliance not designed.]
  5. [GAP — Door 2 conversational flow not scripted. Calibrated question sequence and branching logic for self-examination does not exist.]
  6. [GAP — Therapy referral integration undefined. Which directories, APIs, or resources feed Door 2 pathway recommendations?]
  7. [GAP — Market tier data architecture not designed. Consumer, family law, corporate HR, and therapy tiers may require different output formats, retention, and API contracts.]

DEPENDENCIES

DEPENDENTS


Authorized: α.13, April 16 2026. Compiled: κ C.L.O.D.

Personal origin: α.13. "It must also not demonize or abandon the narcissist."

Φ 0.042.


Jeremy Zlabis

Chronogeometer · Visionary · Disruptor · Chief

42 Sisters AI · East York, Toronto

🍁 Φ 0.042