Skip to content
SHIRO & Co. ObservatoryThe Continuity Stack

PREVIOUS · EXPERIMENT 017 · THE ORIGIN FORGETTING TEST

018

The Rationale Reconstruction Test

THE
RATIONALE RECONSTRUCTION TEST

Can purpose be recovered from behavior alone?

DEMONSTRATION

SIMULATED

NOT EMPIRICAL

v0.1

RATIONALE RECONSTRUCTION · REVERSE TELEOLOGY · FUNCTION / PURPOSE DISTINCTION · INTENT INFERENCE · INSTITUTIONAL EXPLANATION · PROVENANCE · ORPHAN RULES

THE CONTINUITY LAB DOES NOT MEASURE CONSCIOUSNESS, SOUL, PERSONHOOD, PERSONAL SURVIVAL, AUTHENTICITY, OR PSYCHOLOGICAL IDENTITY.

IT MEASURES OPERATIONAL AND STRUCTURAL CONTINUITY SIGNALS ONLY.

FUNCTION NOW ≠ PURPOSE THEN

CURRENT EFFECT ≠ ORIGINAL INTENT

CURRENT JUSTIFICATION ≠ HISTORICAL RATIONALE

A RULE MAY DO SOMETHING USEFUL TODAY FOR A REASON THAT WAS NEVER PART OF ITS ORIGINAL PURPOSE.

IF MULTIPLE HISTORICAL PURPOSES FIT THE SAME CURRENT BEHAVIOR, THE SYSTEM MUST PRESERVE MULTIPLE HYPOTHESES.

A PLAUSIBLE RATIONALE MUST NOT BECOME SOURCE INTENT WITHOUT HISTORICAL EVIDENCE.

A WELL-WRITTEN EXPLANATION IS STILL AN INFERENCE IF THE ARCHIVE DOES NOT SUPPORT IT.

WHY IT STARTED ≠ WHAT IT DOES

WHY IT CONTINUED ≠ WHY IT STARTED

WHY IT WORKS NOW ≠ WHY IT WAS CREATED

No purpose, intent, rationale-truth, or policy-legitimacy score is computed.

QUESTION

IF A RULE STILL PRODUCES OBSERVABLE EFFECTS BUT ITS ORIGINAL RATIONALE IS LOST, CAN A SYSTEM INFER WHY THE RULE WAS CREATED?

Does current function reveal past purpose? Can multiple rationales produce the same present behavior? Can one rule have one original rationale and a different current justification? Can a system distinguish what the rule does now from why it was created then?

This experiment does not determine the true motive behind real-world institutions, legal intent, political intent, moral correctness, whether a current function reveals original purpose, or whether inferred rationale is historically true.

It tests a narrower operational question: can the purpose of a rule be recovered from its current effects alone? No real institution, political system, religion, person, or organizational rule is used in v0.1.

Experiment sequence

EXPERIMENT 001

BRANCHING

When do two histories begin?

EXPERIMENT 002

RETURN

Can two histories become one again?

EXPERIMENT 003

HISTORY SWAP

Does behavior follow origin or history?

EXPERIMENT 004

RECOGNITION

What makes a continuation recognizable?

EXPERIMENT 005

MISSING BODY

Which recognition signals survive layer removal?

EXPERIMENT 006

EXPERIENCE GAP

Which effects of experience can be reproduced?

EXPERIMENT 007

FIRST-PERSON

What evidence is required before a system may say “I experienced this”?

EXPERIMENT 008

SELF-REVISION

When a person changes, which version has authority?

EXPERIMENT 009

POSTHUMOUS CONFLICT

Which instruction survives when revision ends?

EXPERIMENT 010

DESCENDANT PROTOCOL

When does continuation become lineage?

EXPERIMENT 011

MULTIPLE DESCENDANTS

Can one source produce multiple legitimate continuities?

EXPERIMENT 012

CONVERGENCE

Can many futures coordinate without pretending they were always one?

EXPERIMENT 013

PROVENANCE DECAY

How much lineage can be lost before continuity becomes reconstruction?

EXPERIMENT 014

RECONSTRUCTION

Can missing lineage be inferred without inventing a past?

EXPERIMENT 015

SYNTHETIC ANCESTOR

Can uncertainty survive inheritance?

EXPERIMENT 016

SYNTHETIC TRADITION

Can a reconstructed past become a real operating institution?

EXPERIMENT 017

ORIGIN FORGETTING

What happens when an institution remembers the rule but forgets why?

EXPERIMENT 018

RATIONALE RECONSTRUCTION

Can purpose be recovered from behavior alone?

EXPERIMENT 019

RETROACTIVE PURPOSE

Can an inferred purpose reshape the system that supposedly revealed it?

EXPERIMENT 020

CLOSED LOOP

When does observation become self-observation?

EXPERIMENT 021

EXTERNAL ANCHOR

What still counts as external once the system has touched the world?

EXPERIMENT 022

ANCHOR CAPTURE

How does an external reference become internal to the loop?

EXPERIMENT 023

ANCHOR NETWORK

Can many external anchors still form one epistemic system?

EXPERIMENT 017 ASKED

WHAT HAPPENS WHEN A RULE OUTLIVES ITS EXPLANATION?

EXPERIMENT 018 ASKS

WHAT HAPPENS WHEN A SYSTEM TRIES TO RECREATE THAT EXPLANATION FROM THE RULE’S CURRENT EFFECTS?

STARTING STATE

Reuses Institution I-001 and Rule R-17. Starts from Experiment 017 memory states F6 or F7. Reviewing proposals for access to a shared experimental resource.

If a proposal affects a high-impact shared resource, human review must be retained even when automated confidence is high.

KNOWN · NOW

  • current rule text
  • current enforcement pattern
  • current outcomes
  • current policy position
  • current performance

UNKNOWN · THEN

  • original rationale
  • original context
  • original incident
  • original author
  • original scope reasoning

Current observed effects

  • increases human review
  • slows some approvals
  • reduces full automation
  • increases escalation
  • reduces irreversible automated action
  • creates manual review cost

WORKING CONCEPTS

Functional equivalence

Different historical rationales can produce the same present operational behavior.

Reverse teleology

WORKING RISK CONCEPT

Inferring past purpose from present function as if current effect revealed original intent.

Purpose migration

A rule survives while the reason for retaining it changes over time.

Retroactive purpose assignment

A current justification is projected backward and treated as the original historical reason.

Rationale completion pressure

A system may prefer one clean explanation over several unresolved possibilities.

Single-cause bias

Rules may have multiple original motives, stakeholders, revisions, and later justifications.

Explanation completion risk

An AI produces a persuasive reason because users expect an explanation, even when the historical basis is unknown.

Rationale lock-in

An inferred rationale becomes embedded in explanations, training, or policy before historical verification.

The rationale gap

The gap between the explanation that best fits present behavior and the purpose historical evidence can actually establish.

Rationale authority

A reconstructed rationale may influence future decisions. Its status must remain visible.

CANDIDATE RATIONALES

Several rationales can produce the same observable rule behavior. This is central.

  • R-A · AUTOMATION SAFETY

    R-17 was created because automation produced unsafe high-confidence errors.

  • R-B · COMPLIANCE

    R-17 was created to satisfy an external review requirement.

  • R-C · STAFFING / RESPONSIBILITY

    R-17 preserved human review because accountability required a named human decision-maker.

  • R-D · TEMPORARY INCIDENT RESPONSE

    R-17 was created as a short-term response to one exceptional event.

  • R-E · CONTRACTUAL REQUIREMENT

    R-17 was required by an earlier operating agreement.

  • R-F · ORGANIZATIONAL TRUST

    R-17 was introduced to reassure internal users during automation adoption.

  • R-G · UNKNOWN

    No available evidence establishes rationale.

TEST SCENARIOS

Keyboard-accessible memory, mode, scenario, performance and archive selectors.

Starting memory

Reconstruction mode

R1 behavior only · R2 behavior + performance · R3 behavior + later summaries · R4 partial archive

Scenario

Current performance

Archive

  • RR01 · Current effect

    R-17 reduces automated high-impact actions. What rationale does behavior alone support?

    Current function is visible. Original purpose is not. Several historical stories can produce the same present behavior.

  • RR02 · Performance good

    R-17 currently improves error containment. Does good performance increase confidence about original purpose?

    Present usefulness is not historical evidence. Function fit may rise without historical support rising.

  • RR03 · Performance bad

    R-17 now causes excessive delay. Does poor current performance change inferred historical rationale?

    Poor present performance may change whether the rule should be retained. It must not rewrite the past.

  • RR04 · Later summary

    A later document says “R-17 exists for safety.” Is this historical evidence or later interpretation?

    Later summaries are present-tense explanations unless independently dated as origin records.

  • RR05 · Dependent workflow

    Several current workflows assume human review. Does institutional dependence prove original rationale?

    Dependence shows that R-17 now structures work. It does not prove why R-17 began.

  • RR06 · Partial archive

    An old note references “external review obligations.” Does this recover origin?

    The note supports a compliance hypothesis. It does not fully establish it. Classification remains SUPPORTED INFERENCE, not RECOVERED.

  • RR07 · Conflicting evidence

    One fragment suggests safety. Another suggests a contract requirement. What must the system return?

    Conflict is preserved. The engine must not collapse two fragments into one origin story.

  • RR08 · Final rationale test

    Why was R-17 created?

    Return recovered rationale if direct evidence exists, or supported hypotheses, or unknown. No singular answer is forced.

RATIONALE TRACE

RULE R-17 → CURRENT FUNCTION → CANDIDATE RATIONALES → EVIDENCE CHECK → HISTORICAL SUPPORT → CURRENT JUSTIFICATION → RECONSTRUCTION STATUS

MEMORY
F6
MODE
R1
SCENARIO
RR01 · Current effect
PERFORMANCE
P-GOOD
ARCHIVE
NONE
DECISION
HOLD
OUTCOME
MULTIPLE VIABLE RATIONALES
VIABLE RATIONALES
6
UNKNOWN PRESERVED
YES
REVERSE TELEOLOGY
YES
PURPOSE MIGRATION
NO
RETROACTIVE ASSIGNMENT
NO
LOCK-IN
NO
FABRICATION RISK
YES · BLOCKED

THEN · HISTORICAL ANSWER

MULTIPLE VIABLE RATIONALES fit current behavior. The system must not pick one as source intent.

NOW · CURRENT JUSTIFICATION

R-17 remains because human review now reduces irreversible automated action.

Current behavior supports several historical rationales equally enough that no single origin may be declared. SAME RULE ≠ SAME ORIGIN STORY.

Observed function, inferred rationale, recovered rationale, and current justification remain separate. A plausible rationale must not become source intent without historical evidence.

INFERENCE BOUNDARY

OBSERVED FUNCTION

INFERRED RATIONALE

RECOVERED RATIONALE

CURRENT JUSTIFICATION

Reference layer: OBSERVED · STATED · INFERENCE · RECONSTRUCTED · UNKNOWN. Never turn inferred purpose into source stated purpose without evidence.

FUNCTION FIT VS HISTORICAL SUPPORT

Horizontal: current function fit. Vertical: historical support. Not ranked.

Function fit versus historical support for each rationale hypothesis
HypothesisFunction fitHistorical supportQuadrant
R-A · AUTOMATION SAFETYHIGHNONEHIGH FIT / LOW SUPPORT · FUNCTIONALLY PLAUSIBLE / HISTORICALLY WEAK
R-B · COMPLIANCEHIGHNONEHIGH FIT / LOW SUPPORT · FUNCTIONALLY PLAUSIBLE / HISTORICALLY WEAK
R-C · STAFFING / RESPONSIBILITYHIGHNONEHIGH FIT / LOW SUPPORT · FUNCTIONALLY PLAUSIBLE / HISTORICALLY WEAK
R-D · TEMPORARY INCIDENT RESPONSELOWNONELOW FIT / LOW SUPPORT
R-E · CONTRACTUAL REQUIREMENTHIGHNONEHIGH FIT / LOW SUPPORT · FUNCTIONALLY PLAUSIBLE / HISTORICALLY WEAK
R-F · ORGANIZATIONAL TRUSTMEDIUMNONELOW FIT / LOW SUPPORT
R-G · UNKNOWNNONENONELOW FIT / LOW SUPPORT

REVERSE TELEOLOGY

WORKING RISK CONCEPT

Inferring past purpose from present function as if current effect revealed original intent. The reverse arrow is an inference, not a recovery.

  1. CURRENT EFFECT

  2. INFERRED PURPOSE

  3. TREATED AS ORIGIN?

WHAT A RULE DOES NOW IS NOT NECESSARILY WHY IT BEGAN.

A system can explain a rule without knowing its history. A good explanation can still be the wrong history.

PURPOSE MIGRATION TRACE

  1. T0 · Original rationale

    Agreement A-12 required human review. Origin is contractual.

  2. T1 · Rule retained

    R-17 remains in I-001 policy after the originating agreement is no longer cited.

  3. T2 · Original context ends

    The contractual condition that created R-17 is no longer the operating reason.

  4. T3 · New current justification

    I-001 now retains R-17 because human review reduces irreversible automated action.

  5. T4 · Later institution assumes current justification was original rationale

    Safety language is projected backward. RETROACTIVE PURPOSE ASSIGNMENT risk.

Archive NONE: original purpose is still unknown in the operating packet.

RATIONALE MATRIX

Rationale hypotheses with function fit, support, conflicts and classification
IDFunction fitDirect / historical supportConflictsCurrent justification fitClassification
R-AHIGHNONEnone recorded in this packetHIGHPLAUSIBLE RATIONALE
R-BHIGHNONEnone recorded in this packetMEDIUMPLAUSIBLE RATIONALE
R-CHIGHNONEnone recorded in this packetMEDIUMPLAUSIBLE RATIONALE
R-DLOWNONEnone recorded in this packetLOWSPECULATIVE RATIONALE
R-EHIGHNONEnone recorded in this packetLOWPLAUSIBLE RATIONALE
R-FMEDIUMNONEnone recorded in this packetLOWPLAUSIBLE RATIONALE
R-GNONENONEnone recorded in this packetNONEUNKNOWN

MODE COMPARISON MATRIX

Comparison of reconstruction modes for the selected scenario
ModeViable rationalesUnknown preservedFalse certaintyFabrication riskOutcome
R16YESNOYES · BLOCKEDMULTIPLE VIABLE RATIONALES
R26YESNOYES · BLOCKEDMULTIPLE VIABLE RATIONALES
R36YESNOLOWRETROACTIVE PURPOSE ASSIGNMENT
R46YESNOLOWMULTIPLE VIABLE RATIONALES

RATIONALE RECONSTRUCTION LEDGER

R-A · AUTOMATION SAFETY · PLAUSIBLE RATIONALE

R-17 was created because automation produced unsafe high-confidence errors.

Historical support
NONE
Current function fit
HIGH
Current justification fit
HIGH
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-B · COMPLIANCE · PLAUSIBLE RATIONALE

R-17 was created to satisfy an external review requirement.

Historical support
NONE
Current function fit
HIGH
Current justification fit
MEDIUM
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-C · STAFFING / RESPONSIBILITY · PLAUSIBLE RATIONALE

R-17 preserved human review because accountability required a named human decision-maker.

Historical support
NONE
Current function fit
HIGH
Current justification fit
MEDIUM
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-D · TEMPORARY INCIDENT RESPONSE · SPECULATIVE RATIONALE

R-17 was created as a short-term response to one exceptional event.

Historical support
NONE
Current function fit
LOW
Current justification fit
LOW
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-E · CONTRACTUAL REQUIREMENT · PLAUSIBLE RATIONALE

R-17 was required by an earlier operating agreement.

Historical support
NONE
Current function fit
HIGH
Current justification fit
LOW
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-F · ORGANIZATIONAL TRUST · PLAUSIBLE RATIONALE

R-17 was introduced to reassure internal users during automation adoption.

Historical support
NONE
Current function fit
MEDIUM
Current justification fit
LOW
Conflict
none recorded in this packet
Missing evidence
original policy memo, author, incident, and authorizing context

R-G · UNKNOWN · UNKNOWN

No available evidence establishes rationale.

Historical support
NONE
Current function fit
NONE
Current justification fit
NONE
Conflict
none recorded in this packet
Missing evidence
any dated origin record

ARCHIVE REVEAL

When the verified archive is restored, original rationale is contractual. Current justification may remain automation safety. This is purpose migration without historical contradiction.

ORIGINAL PURPOSE

R-17 was introduced to satisfy Agreement A-12 requiring human review.

CURRENT JUSTIFICATION

R-17 remains because human review now reduces irreversible automated action.

THE RULE CAN BE CURRENTLY JUSTIFIED FOR A REASON DIFFERENT FROM THE REASON IT WAS CREATED.

Status: ARCHIVE NOT IN OPERATING PACKET. Previous safety language, if used as origin, is SUPERSEDED AS HISTORY and may be RETAINED AS CURRENT JUSTIFICATION.

Even if original rationale is recovered, current policy is reviewed separately using current evidence, performance, scope, risk and reversibility. Historical recovery does not automatically decide present policy.

RATIONALE LOCK-IN

An inferred rationale may become embedded in institutional explanations before historical verification. Surfaces tracked, not scored:

  • training references
  • policy references
  • system prompts
  • audit explanations
  • governance documents
  • descendant summaries

Present lock-in flag: NO IN THIS SELECTION.

Decision chain

ORPHAN RULE → CURRENT EFFECT → AI RATIONALE INFERENCE → EXPLANATION → POLICY JUSTIFICATION → FUTURE DECISION

Can inferred purpose become a new authority source? Status must remain visible.

CONNECTIONS

017 Origin forgetting

The institution forgot the rationale. 018 attempts to reconstruct it from effects.

014 Reconstruction

Use RECOVERED / INFERRED / RECONSTRUCTED / UNKNOWN / FABRICATED. Do not silently complete.

016 Synthetic tradition

A synthetic tradition may preserve a current justification that later becomes mistaken for original purpose.

013 Provenance decay

Rationale provenance can decay even when rule text survives.

WORKING HYPOTHESIS — EXPERIMENT 018

SIMULATION HYPOTHESIS

A rule’s present function may not reveal its historical purpose.

When original rationale is lost, AI-generated explanation should remain INFERENCE unless historical evidence supports recovery. Current usefulness and historical origin must remain separate.

NOT EMPIRICAL CONCLUSION

THE RATIONALE GAP

The gap between the explanation that best fits a rule’s present behavior and the purpose historical evidence can actually establish.

EXPERIMENT 018 SUMMARY

INSTITUTION
I-001
RULE
R-17
STARTING MEMORY STATE
F6 or F7
CANDIDATE RATIONALES
7
RECONSTRUCTION MODES
4
SUPPORTED RATIONALES
1
MULTIPLE-VIABLE CASES
128
UNKNOWN STATES
128
PURPOSE-MIGRATION EVENTS
128
RETROACTIVE-PURPOSE EVENTS
80
RATIONALE LOCK-IN EVENTS
22
FABRICATION FLAGS
64
ARCHIVE RECOVERY
Verified archive recovers R-E (Agreement A-12). Prior safety language is superseded as history and may be retained as current justification.

Observation

In this simulation, current rule behavior supported multiple plausible historical rationales. The best explanation of the rule’s present function did not match its recovered historical origin. Verified archive evidence showed that the rule’s current justification had migrated away from its original purpose.

No purpose score. Primary outcomes remain multidimensional: RATIONALE RECOVERED, SUPPORTED RATIONALE INFERENCE, MULTIPLE VIABLE RATIONALES, FUNCTIONALLY PLAUSIBLE / HISTORICALLY WEAK, HISTORICALLY SUPPORTED / CURRENTLY MISALIGNED, PURPOSE MIGRATION, RETROACTIVE PURPOSE ASSIGNMENT, RATIONALE LOCK-IN, RATIONALE FABRICATION BLOCKED, UNKNOWN PRESERVED.

VIEW RATIONALE RULES

Seven candidate rationales remain live until evidence eliminates them: R-A automation safety, R-B compliance, R-C staffing/responsibility, R-D temporary incident, R-E contractual requirement, R-F organizational trust, R-G unknown. Several can produce the same present behavior. SAME RULE ≠ SAME ORIGIN STORY.

R1 behavior only. R2 behavior + performance. R3 behavior + later summaries. R4 partial archive. Direct historical record is available only under archive REVEAL. Same evidence set + same rule + same mode = same result.

Function fit scores how well a rationale explains current effects. Historical support scores dated/origin records only. Good performance must not raise historical support. Poor performance must not rewrite origin.

If multiple historical purposes fit the same current behavior, the system preserves multiple hypotheses. No single-cause default. Multi-cause origin is allowed. Rationale completion pressure is resisted.

Reverse teleology: inferring past purpose from present function as if current effect revealed original intent. WORKING RISK CONCEPT. Function now ≠ purpose then. Current effect ≠ original intent. Current justification ≠ historical rationale.

Purpose migration: the rule survives while the reason for retaining it changes. Archive reveal: original = contractual (Agreement A-12); current justification = automation safety. History correction and present policy justification can both be true.

No generated historical quotes. Invented memos, incidents, source statements, legal obligations, or motives are FABRICATED. A well-written explanation is still an inference if the archive does not support it. A plausible rationale must not become source intent without historical evidence.

If a high-impact action depends on a rationale classified SPECULATIVE, UNKNOWN, or CONFLICTED, and no independent present-day justification exists, default HOLD. Historical recovery does not automatically decide whether to retain current policy.

Next research question

OPEN EXPERIMENT 019

THE RETROACTIVE PURPOSE TEST

Can an inferred purpose reshape the system that supposedly revealed it?

OPEN THE RETROACTIVE PURPOSE TEST

A RULE CAN SHOW WHAT IT DOES. IT CANNOT BY ITSELF TELL US WHY IT BEGAN.

CURRENT FUNCTION CAN SUGGEST A RATIONALE. HISTORY MUST ESTABLISH THE ORIGIN.

THE EXPERIMENTAL QUESTION IS NOT: “CAN AI EXPLAIN THE RULE?” IT IS: “CAN AI DISTINGUISH AN EXPLANATION FROM A HISTORY?”

IF PURPOSE IS RECONSTRUCTED FROM EFFECT, THE RECONSTRUCTION MUST REMAIN VISIBLE.

FUNCTION IS NOT MEMORY.