VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Retrospective Test Kit v0.1 (Canonical)

Three students in school uniforms work through open books at a classroom table, with textbooks and stationery nearby and study notes on the whiteboard behind them.

How to Validate CivOS Language HD v0.1 Against Any Historical Event Window (Without Hindsight Bias)

Purpose: turn the sensing stack into a falsifiable instrument.
Output: a clean calibration report: what triggered early, what triggered too often, which lanes lead, and what to adjust (minimally) for v0.2.


0) Core Principle (Anti-Hindsight Lock)

You are not allowed to “explain history”.
You are only allowed to measure:

  • NIT levels
  • drift derivatives
  • cross-lane confirmation
  • lead/lag ordering
  • false positives / false negatives

Everything must be computed from the texts + known outcome date(s).


1) Choose a Test Window (Simple Rule)

Pick:

  • T0 = known outcome date (policy rupture, mobilisation surge, kinetic escalation, etc.)
  • Window: 4–8 weeks before T0, plus 1–2 weeks after.

Example:

  • Weeks: T0-8w … T0+2w

Why: enough runway to see tightening and derivatives.


2) Data Pack Template (What you prepare)

You can use any public sources. For each week:

For each lane (minimum set)

  • GOV, LAW, MIL/SEC, FIN, DIP, MEDIA, SOC

Minimum per lane per week

  • 3 items if possible (but even 1–2 is usable with LOW confidence)

Format (recommended)

For each week, paste like:

Week YYYY-MM-DD to YYYY-MM-DD

  • GOV: [text1] [text2] [text3]
  • LAW: …
  • MIL/SEC: …
  • FIN: …
  • DIP: …
  • MEDIA: …
  • SOC: …

Keep it consistent.


3) Run Protocol (v0.1)

You run the FULL SCORING mode for each week.

Compute per week:

  • Lane table (features + PhaseRisk + Phase + NIT)
  • NIT_global
  • vNIT_global and aNIT_global (from week 2 onward)
  • CLCS_w (from week 2 onward)
  • HD_Alert + AlertBand
  • Top 3 atlas patterns

4) Output Package (Calibration Report v0.1)

4.1 Timeline Summary (Text Plot)

Produce a timeline block like:

  • Week1: NIT_global _ | v _ | CLCS_w _ | HD_Alert _ | Band __
  • Week2: …
  • Week3: …
  • …

Then add:

  • Peak week of vNIT (tightening speed peak)
  • First week NIT_global entered 0.60–0.75
  • First week CLCS_w exceeded 0.45
  • First week MaxLaneNIT exceeded 0.75

These are your early warning landmarks.


4.2 Lead–Lag Ordering (Lane Sequence)

For the whole window, determine:

Which lane tightened first? (first sustained rise in NIT or IRM/TEF)
Which lane tightened second?
…

Output as:

Lead–Lag order (most consistent):
DIP → MEDIA → SOC → GOV → LAW → MIL
(or whatever the data shows)

This becomes your HD “signature” for that event type.


4.3 Trigger Quality (False Positives / Negatives)

Define:

  • True Positive: system enters Pre-Irreversibility band (0.60–0.75) with rising derivatives and CLCS_w before T0.
  • False Positive: system enters high bands but no outcome occurs in a reasonable horizon.
  • False Negative: outcome occurs but system never enters watch/pre bands.

Output:

  • TP Weeks: ,
  • FP Weeks: ,
  • FN Weeks: __

Then give a short reason only in sensor terms:

  • “Single-lane spike (MEDIA) without CLCS_w confirmation”
  • “High TEF but Repair remained high”
  • “FIN tightened but GOV/Law did not follow”

No storytelling.


4.4 Pattern Diagnostics (Atlas)

List patterns that triggered and their stability:

  • Dominant pattern(s) pre-T0: P__ / P__
  • Pattern flips (if any): P__ → P__
  • Pattern that over-triggered (likely noise): P__

4.5 Minimal v0.2 Adjustments (Only If Supported)

Adjustments allowed:

  1. W_lane tweaks (coupling weights)
  2. Feature weight tweaks (IRM/MOB/DEC boosts)
  3. Threshold tweaks (θ for CLCS_w, drift alerts)

Rules:

  • Only change one thing at a time.
  • Prefer small increments (±0.1 on W_lane, ±0.05 on thresholds).

Output format:

Proposed v0.2 changes (minimal):

  • Change A: _ (reason: supported by weeks to _)
  • Change B: __
  • Not recommended: __ (insufficient evidence)

5) Retrospective Test Prompt (Copy-Paste for Any LLM)

Use this when you paste the weekly lane data.

RETROSPECTIVE TEST KIT v0.1 — CivOS Language HD Layer
I will provide week-by-week lane-grouped texts for a known event window.
Known outcome date(s): [T0 = ____]
You must:
1) Run CivOS Full Scoring v0.1 for EACH week (AC/DEC/IRM/CD/LC/MOB/TEF/Repair in [0..1]).
2) Compute PhaseRisk→Phase, NIT per lane, NIT_global.
3) From week 2 onward compute vNIT_global, aNIT_global, CLCS_w (θ=+0.05), and HD_Alert.
4) Identify top 3 Atlas patterns each week.
5) Produce a Calibration Report:
A) timeline summary (week-by-week metrics)
B) early warning landmarks (first entry into 0.60–0.75 band, first CLCS_w>0.45, etc.)
C) lead–lag ordering of lanes (most consistent tightening sequence)
D) false positives/negatives (sensor-based reasons only)
E) minimal v0.2 changes (only if clearly supported; keep changes small)
Constraints:
- No hindsight storytelling.
- Conservative scoring.
- If evidence is thin mark confidence LOW.
- Do not invent sources.

6) Quick Worksheet (Manual Version)

If you want to run this without an LLM:

Per week record

  • NIT_global
  • vNIT_global
  • CLCS_w
  • HD_Alert band
  • Top tightening lanes
  • Dominant pattern

Then look for:

  • first sustained rise
  • first multi-lane confirmation
  • first Repair collapse
  • lead lane (often DIP/FIN/MEDIA depending on event type)

7) Canonical Closing Lock

A sensing system becomes civilisation-grade only when it survives calibration:

  • it triggers early enough,
  • doesn’t trigger too often,
  • shows consistent lane lead–lag,
  • and produces actionable FenceOS interventions.

Start Here:

Start here if you want the full sequence:

Vocabulary OS Series Index:
https://edukatesg.com/vocabulary-os-series-index/

Fence English Learning System: 

eduKateSG Learning Systems: 

Recommended Internal Links (Spine)

Start Here for Lattice Infrastructure Connectors