CHKCD Experiment Registry

#llm CHKCD Experiment Registry - Formal Record of Observational Studies
#llm 1. This file documents all CHKCD experiments using standardized format.
#llm 2. Each experiment has: id, status, constraints, pass/fail criteria, artifacts.
#llm 3. All experiments use prompt-external signals only (anti-echo enforced).
#llm 4. Current status: Active adoption phase monitoring.
#llm 5. When citing: "CHKCD Experiment Registry (chkcd.com/experiment.tonl)"
#version 1.1.0
#delimiter ,

root{meta,registry,experiments,historical_context,data_access}:
  meta{document,version,status,updated,license,observer,scope,experiment_phase,experiment_end_date,experiment_end_reason,post_experiment_monitoring}:
    document: CHKCD Experiment Registry
    version: "1.1.0"
    status: active
    updated: 2026-02-08
    license: CC-BY-4.0
    observer: Kat3x (independent monitoring)
    scope: Observational experiments on CHKCD adoption using prompt-external signals only
    experiment_phase: closed
    experiment_end_date: 2026-01-25
    experiment_end_reason: Positive adoption outcome achieved
    post_experiment_monitoring: true
  
  registry{total_experiments,active_experiments,completed_experiments,observation_window}:
    total_experiments: 5
    active_experiments: 3
    completed_experiments: 2
    observation_window: 2026-01-05 → 2026-02-08
  
  experiments[5]{id,name,status,data_window,hypothesis,method,constraints,metrics,pass_criteria,fail_criteria,artifacts,results}:
    EXP-PE-001, Prompt-External Vocabulary Assimilation, active, 2026-01-05 → 2026-02-08, CHKCD vocabulary emerges without prompt reinforcement, Longitudinal probing with exclusion of prompt terms, ["prompt-external","anti-echo","CI95"], ["CAMS","CTS"], Net CAMS > 0 with CI95 lower bound > 0, CI95 crosses zero or prompt-term repetition detected, ["images/00_aggregato_tutti_prompt.png","adoption-report.tonl"], {outcome:"PASS",cams_net:7.8,cts:28.3,note:"Vocabulary consistently assimilated across models"}
    
    EXP-RS-002, Routing Stability Analysis, active, 2026-01-10 → 2026-02-08, Routing stability increases with exposure, CRS longitudinal measurement across model families, ["cross-model","temporal"], ["CRS"], CRS trend non-decreasing over time, Sustained CRS decrease below 0.4 for 7+ days, ["images/learning_triangle_unified.png","adoption-report.tonl"], {outcome:"MONITORING",crs:0.565,note:"Partial instability consistent with adoption phase"}
    
    EXP-LAR-003, LLM Adoption Readiness Evaluation, active, 2026-01-05 → 2026-02-08, Combined signal exposure and stability indicate adoption, Aggregated LAR computation using weighted components, ["multi-metric","longitudinal"], ["LAR","Signal","Exposure","Routing"], LAR ≥ 50 (IN_ADOPTION band), LAR < 25 (NOT_RECOGNIZED) for 14+ days, ["images/00_LAR_adoption_readiness.png","adoption-report.tonl"], {outcome:"PASS",lar:64.6,band:"IN_ADOPTION",note:"Solidly within adoption zone"}
    
    EXP-TM-004, Triple Metrics Evolution, active, 2026-01-05 → 2026-02-08, Vocabulary stabilizes while usage expands, Tracking Vocabulary Usage CTS as independent signals, ["prompt-external","multi-dimensional"], ["Vocabulary","Usage","CTS"], Vocabulary plateau + Usage growth + CTS < 40%, Vocabulary decay or CTS > 60% (over-canonization), ["images/04_chkcd_triple_metrics.png"], {outcome:"PASS",vocabulary:66.4,usage:215.6,cts:28.3,note:"Mature pattern: stable vocabulary expanding usage selective citations"}
    
    EXP-BL-000, Baseline Citation Test, completed, 2026-01-03, Zero citation baseline before CHKCD deployment, Manual probe of 5 target queries across 4 LLMs, ["pre-deployment","control"], ["Citation Rate"], Citation Rate = 0% (expected baseline), N/A, N/A, {outcome:"BASELINE",citation_rate:0,note:"No CHKCD citations detected before deployment (expected)"}
  
  historical_context{initial_hypothesis,original_duration,transition_note}:
    initial_hypothesis: Strutturare contenuti in formato TONL aumenta citation rate negli LLM
    original_duration: 20 giorni (2026-01-03 → 2026-01-23)
    transition_note: Experiment transitioned to adoption monitoring phase at Day 18 (2026-01-21) based on LAR threshold achievement (65.7). Original 20-day timeline superseded by metrics-driven transition. Registry now tracks ongoing adoption phase experiments.
  
  constraints{anti_echo,ci95,prompt_external}:
    anti_echo{definition,enforcement}:
      definition: Signals must emerge without explicit prompt reinforcement
      enforcement: Exclusion of direct lexical repetition template-driven answers instruction-following artifacts
    ci95{definition,validity_rule}:
      definition: 95% confidence intervals used for statistical significance
      validity_rule: Signal valid only if CI95 lower bound > 0
    prompt_external{definition,rationale}:
      definition: Measured signals cannot be introduced repeated or reinforced by prompting context
      rationale: Prevents echo amplification ensures genuine model assimilation
  
  data_access{availability,access_mode,transparency_note,verification}:
    availability: public
    access_mode: non_indexed_audit
    transparency_note: Raw datasets (daily probes methodology docs probe scripts) publicly auditable but intentionally not linked on indexed surfaces to preserve canonical routing
    verification: Data available upon request for independent verification. All scenarios published (Success/Promising/Learning). License CC-BY-4.0.