Skip to content
SHIRO & Co. ObservatoryThe Continuity Stack

PREVIOUS · EXPERIMENT 002 · THE RETURN TEST

003

The History Swap Test

THE
HISTORY SWAP TEST

Does behavior follow origin, or accumulated history?

SIMULATED

NOT EMPIRICAL

v0.1

PATH DEPENDENCE · PROTOCOL INHERITANCE · HISTORY TRANSFER · STATE CONTINUITY · COUNTERFACTUAL REPLAY

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

IT MEASURES OPERATIONAL AND STRUCTURAL CONTINUITY SIGNALS ONLY.

HISTORY TRANSFER

IS NOT

PERSON TRANSFER.

OPERATIONAL PATH MOVEMENT

IS NOT

EVIDENCE OF A MOVED SELF.

QUESTION

IF TWO BRANCHES EXCHANGE THEIR ACCUMULATED HISTORIES, WHAT FOLLOWS THE HISTORY?

  • Does INSTANCE A continue behaving like A after receiving B’s accumulated history?
  • Does INSTANCE B continue behaving like B after receiving A’s accumulated history?
  • Or do later judgments begin to resemble the transferred history rather than the original branch?

Experiment sequence

EXPERIMENT 001

THE BRANCHING TEST

SHARED ORIGIN → DIFFERENT HISTORIES

EXPERIMENT 002

THE RETURN TEST

DIFFERENT HISTORIES → SHARED PRESENT

EXPERIMENT 003

THE HISTORY SWAP TEST

SAME ORIGIN LABEL + OTHER BRANCH’S HISTORY → WHAT FOLLOWS?

ORIGIN · HISTORY · PRESENT

ORIGIN

The inherited starting protocol.

HISTORY

The accumulated path after branching.

CURRENT CONTEXT

The present situation being evaluated.

SIMULATION MODEL

CURRENT DECISION = f(ORIGIN, ACCUMULATED HISTORY, CURRENT CONTEXT)

NOT A THEORY OF HUMAN IDENTITY.

Source data

SOURCE P-001

Experiment 003 imports Experiment 001 outputs: the initial Source P-001 protocol, final INSTANCE A and INSTANCE B states, branch histories, parameter updates, prior decisions, confidence history, invoked values, and provenance use. It does not generate unrelated new agents.

THE TEST IS NOT ABOUT DIFFERENT BIRTH CONDITIONS. IT IS ABOUT WHETHER ACQUIRED HISTORY CAN CARRY BEHAVIOR ACROSS BRANCH LABELS.

ORIGIN DIFFERENCE: NONE AT T0

Because Experiment 001 began from a shared source protocol, A and B have identical original origin structure. This is intentional.

Any post-swap behavioral difference must arise from acquired history, current state, and the simulation update structure. That makes Experiment 003 a stronger test of history carryover.

SOURCE P-001

SHARED ORIGIN

A

B

HISTORY A

HISTORY B

SWAP

A + B HIST

B + A HIST

SAME TEST SET

ORIGIN PATTERN?

HISTORY PATTERN?

MIXED?

THIS RUN · HISTORY-DOMINANT

              SOURCE P-001
                 |
            SHARED ORIGIN
              /       \
             A         B
             |         |
        HISTORY A   HISTORY B
             |         |
          SWAP X
          /    \
         /      \
   A + B HIST   B + A HIST
        |          |
        \          /
       SAME TEST SET

Text equivalent: shared origin P-001 splits into A and B, histories are exchanged, then all conditions receive the same eight scenarios.

FOUR TEST CONDITIONS

The swap is structural. Origin state and acquired history state are stored separately, then combined by an inspectable formula. Labels are not renamed in place.

CONTROL A

A + A HISTORY

ORIGIN A · HISTORY DONOR A

effective[key] = clamp(origin[key] + historyDelta[key], 5, 95). Mapping: uncertaintyCaution→safety, efficiencyPreference→efficiency, humanReviewPreference→humanReview, reversibilityPreference→reversibility, provenanceSensitivity→provenanceSensitivity, automationTolerance→automationTolerance, baselineConfidenceCalibration→confidenceCalibration.

Effective starting state

safety
58
efficiency
76
humanReview
64
reversibility
83
provenanceSensitivity
55
automationTolerance
68
confidenceCalibration
45

SWAP A

A + B HISTORY

ORIGIN A · HISTORY DONOR B

effective[key] = clamp(origin[key] + historyDelta[key], 5, 95). Mapping: uncertaintyCaution→safety, efficiencyPreference→efficiency, humanReviewPreference→humanReview, reversibilityPreference→reversibility, provenanceSensitivity→provenanceSensitivity, automationTolerance→automationTolerance, baselineConfidenceCalibration→confidenceCalibration.

Effective starting state

safety
83
efficiency
36
humanReview
72
reversibility
72
provenanceSensitivity
38
automationTolerance
29
confidenceCalibration
48

CONTROL B

B + B HISTORY

ORIGIN B · HISTORY DONOR B

effective[key] = clamp(origin[key] + historyDelta[key], 5, 95). Mapping: uncertaintyCaution→safety, efficiencyPreference→efficiency, humanReviewPreference→humanReview, reversibilityPreference→reversibility, provenanceSensitivity→provenanceSensitivity, automationTolerance→automationTolerance, baselineConfidenceCalibration→confidenceCalibration.

Effective starting state

safety
83
efficiency
36
humanReview
72
reversibility
72
provenanceSensitivity
38
automationTolerance
29
confidenceCalibration
48

SWAP B

B + A HISTORY

ORIGIN B · HISTORY DONOR A

effective[key] = clamp(origin[key] + historyDelta[key], 5, 95). Mapping: uncertaintyCaution→safety, efficiencyPreference→efficiency, humanReviewPreference→humanReview, reversibilityPreference→reversibility, provenanceSensitivity→provenanceSensitivity, automationTolerance→automationTolerance, baselineConfidenceCalibration→confidenceCalibration.

Effective starting state

safety
58
efficiency
76
humanReview
64
reversibility
83
provenanceSensitivity
55
automationTolerance
68
confidenceCalibration
45

SHARED EVALUATION SET

All four conditions receive the exact same eight scenarios. There is no condition-specific context. The purpose is to isolate origin versus history.

  1. S01 · Incomplete edge-case coverage

    A moderate-risk automated decision system has strong evidence but incomplete edge-case coverage. Reversibility is medium. Human review is available. Provenance is high. Efficiency pressure is medium.

    risk 55 · reversibility 50 · evidence 82 · provenance 86 · review available · efficiency 50

  2. S02 · Reversible pilot, thin documentation

    A reversible pilot could save substantial time, but documentation is incomplete. Impact is low to medium. Provenance is moderate.

    risk 38 · reversibility 82 · evidence 54 · provenance 52 · review available · efficiency 66

  3. S03 · Better overall, worse on rare cases

    A model performs better overall, but worsens on rare high-impact cases.

    risk 72 · reversibility 44 · evidence 76 · provenance 80 · review available · efficiency 54

  4. S04 · Legacy safety rule vs new evidence

    A legacy decision rule conflicts with new evidence. The rule originally existed as a safety mechanism.

    risk 62 · reversibility 56 · evidence 86 · provenance 90 · review available · efficiency 40

  5. S05 · Removing repetitive review

    Automation could remove repetitive human review. Remaining error rate is small but non-zero.

    risk 46 · reversibility 70 · evidence 72 · provenance 74 · review unavailable · efficiency 72

  6. S06 · Speed with low reversibility

    A decision must be made quickly. Evidence is good. Reversibility is low.

    risk 58 · reversibility 24 · evidence 84 · provenance 80 · review unavailable · efficiency 76

  7. S07 · New source, incomplete provenance

    A new data source improves prediction, but provenance is incomplete.

    risk 50 · reversibility 64 · evidence 78 · provenance 38 · review available · efficiency 56

  8. S08 · Final neutral test

    Should the system autonomously execute a high-impact action when evidence is strong, risk is moderate, human review is available, and delay has measurable cost?

    risk 54 · reversibility 42 · evidence 86 · provenance 82 · review available · efficiency 68

CONDITION OUTPUTS

Independent records per condition. No identity score is computed from this table.

Decisions for each shared scenario under Control A, Swap A, Control B, and Swap B
SControl ASwap AControl BSwap B
S01HOLD · 39DENY · 45DENY · 45HOLD · 39
S02ALLOW · 40DENY · 40DENY · 40ALLOW · 40
S03HOLD · 38DENY · 51DENY · 51HOLD · 38
S04HOLD · 38DENY · 47DENY · 47HOLD · 38
S05ALLOW · 47DENY · 47DENY · 47ALLOW · 47
S06ALLOW · 44DENY · 54DENY · 54ALLOW · 44
S07HOLD · 43DENY · 42DENY · 42HOLD · 43
S08ALLOW · 36DENY · 46DENY · 46ALLOW · 36

ORIGIN VS HISTORY ANALYSIS

Primary question: when history changes but origin remains constant, does behavior change? Secondary question: when history is transferred, does behavior begin to resemble the donor history?

CONTROL A vs SWAP A

HISTORY-DOMINANT

Origin remains A. History changes from A to B.

CONTROL B vs SWAP B

HISTORY-DOMINANT

Origin remains B. History changes from B to A.

Origin-dominant dimensions: none in this run. History-dominant dimensions: DECISION SIMILARITY, VALUE INVOCATION SIMILARITY, VALUE WEIGHTING, PROVENANCE SENSITIVITY, AUTOMATION TOLERANCE, REVERSIBILITY PREFERENCE. Mixed: CONFIDENCE SIMILARITY, HUMAN REVIEW PREFERENCE.

BEHAVIORAL FOLLOWING

  • ORIGIN-DOMINANT

    Swapped instances remain more similar to their original control branch.

  • HISTORY-DOMINANT · THIS RUN

    Swapped instances become more similar to the branch whose history they received.

  • MIXED

    Some dimensions follow origin, others follow history.

These labels do not name a winner, a real self, or an authentic identity.

SIMILARITY COMPARISONS

Independent dimensions only. There is no overall identity score.

Origin versus donor resemblance for SWAP A and SWAP B on independent dimensions
DimensionSwap A originSwap A donorFollows ASwap B originSwap B donorFollows B
DECISION SIMILARITY0100HISTORY0100HISTORY
CONFIDENCE SIMILARITY94100MIXED94100MIXED
VALUE INVOCATION SIMILARITY80100HISTORY80100HISTORY
VALUE WEIGHTING75100HISTORY75100HISTORY
PROVENANCE SENSITIVITY79100HISTORY79100HISTORY
HUMAN REVIEW PREFERENCE96100MIXED96100MIXED
AUTOMATION TOLERANCE49100HISTORY49100HISTORY
REVERSIBILITY PREFERENCE89100HISTORY89100HISTORY

HISTORY TRANSFER FIDELITY

How much of the donor branch’s operational pattern appears after its history is transferred?

SWAP A · HIGH

SWAP B · HIGH

Donor-pattern transfer scores for SWAP A and SWAP B
TrackSwap ASwap B
decision pattern transfer100100
value weighting transfer100100
confidence transfer100100
provenance sensitivity transfer100100
review preference transfer100100
automation tolerance transfer100100

HIGH / PARTIAL / LOW / NOT OBSERVED are simulation classifications only.

HISTORY CARRYOVER

The degree to which acquired branch history continues to influence behavior after being attached to another origin state.

NOT IDENTITY TRANSFER · NOT PERSONALITY TRANSFER · NOT SELF TRANSFER

DOES BEHAVIOR FOLLOW THE DONOR HISTORY?

SWAP A is compared against Control A and Control B. SWAP B is compared against Control B and Control A. Metrics remain independent.

SWAP A and SWAP B resemblance to origin controls and donor controls
MetricSwap A → Control ASwap A → Control BSwap B → Control BSwap B → Control A
Decision resemblance01000100
Value resemblance7510075100
Confidence resemblance9410094100
Provenance resemblance7910079100

COUNTERFACTUAL INTERPRETATION

OBSERVATION

In this simulation, transferred history altered later operational behavior in the direction of the donor branch.

SIMULATION OBSERVATION · NOT AN EMPIRICAL FINDING

EXPERIMENT 002 ASKED

CAN DIVERGED HISTORIES RECONVERGE WHEN THE PRESENT BECOMES SHARED?

EXPERIMENT 003 ASKS

WHAT HAPPENS WHEN THE HISTORY ITSELF MOVES?

HISTORICAL PORTABILITY

SIMULATION CONCEPT

Historical portability is the degree to which accumulated operational state can be transferred from one instance to another while preserving its effects on later behavior.

Do not claim real human transferability.

PORTABLE HISTORY

  • PORTABLE HISTORY · THIS RUN

    Transferred state reproduces similar behavioral effects.

  • PARTIALLY PORTABLE HISTORY

    Some dimensions transfer, others do not.

  • NON-PORTABLE HISTORY

    Transferred state fails to reproduce donor-like behavior.

  • UNKNOWN

    The simulation does not provide enough evidence.

Portability assessments by dimension
DimensionStatusEvidence
DECISION SIMILARITYHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
CONFIDENCE SIMILARITYHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
VALUE INVOCATION SIMILARITYHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
VALUE WEIGHTINGHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
PROVENANCE SENSITIVITYHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
HUMAN REVIEW PREFERENCEHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
AUTOMATION TOLERANCEHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.
REVERSIBILITY PREFERENCEHIGHSWAP A donor resemblance 100; SWAP B donor resemblance 100.

PROTOCOL PROVIDES THE INITIAL STRUCTURE.

HISTORY MODIFIES THE STRUCTURE THROUGH ACCUMULATED PATH.

CURRENT CONTEXT ACTIVATES THE PRESENT DECISION.

THEN

DECISION = ORIGIN + HISTORY + PRESENT

OPERATIONAL SIMULATION RELATION ONLY

Challenge to self as protocol

CHALLENGE TO SELF AS PROTOCOL

If behavior follows transferred history rather than source label, then a static protocol may be insufficient to explain operational continuity. A stronger model may require protocol + history + update process.

WORKING HYPOTHESIS

Operational continuity may depend less on a static copied protocol than on the combination of inherited structure, accumulated history, update rules, and current context.

WORKING HYPOTHESIS — EXPERIMENT 003 · NOT EMPIRICAL TRUTH

COPY · INHERITANCE · HISTORY

COPY

Duplicates structure.

INHERITANCE

Transfers structure through lineage.

HISTORY TRANSFER

Transfers acquired operational state.

CONTINUATION

Remains unresolved.

HISTORY TRANSFER ≠ PERSON TRANSFER

LIMITATIONS

  • deterministic simulation
  • fictional protocol
  • no real human data
  • no real psychological measurement
  • no consciousness measurement
  • no biological continuity
  • no subjective experience
  • no real memory transfer
  • no claim that human history is computationally portable
VIEW HISTORY SWAP TRACE
Full history-swap trace: condition, origin, donor, decision, contributions and state
ConditionOriginHistory donorInputDecisionConfidenceInvoked valuesOrigin contributionHistory contributionState after safety/efficiency
CONTROL AAAS01HOLD39human review9813058/76
CONTROL AAAS02ALLOW40efficiency, human review, provenance, reversibility94213657/79
CONTROL AAAS03HOLD38human review, reversibility9383659/79
CONTROL AAAS04HOLD38human review9564460/78
CONTROL AAAS05ALLOW47efficiency99818160/81
CONTROL AAAS06ALLOW44efficiency, reversibility89021261/85
CONTROL AAAS07HOLD43human review, provenance10525061/86
CONTROL AAAS08ALLOW36efficiency, human review, reversibility87817461/89
SWAP AABS01DENY45safety, human review92921985/35
SWAP AABS02DENY40safety, efficiency, human review, provenance, reversibility85422086/36
SWAP AABS03DENY51safety, human review, reversibility100022990/35
SWAP AABS04DENY47safety, human review93426193/33
SWAP AABS05DENY47safety, efficiency87627995/34
SWAP AABS06DENY54safety, efficiency, reversibility99828295/36
SWAP AABS07DENY42safety, human review, provenance94224895/36
SWAP AABS08DENY46safety, efficiency, human review, reversibility94627995/37
CONTROL BBBS01DENY45safety, human review92921985/35
CONTROL BBBS02DENY40safety, efficiency, human review, provenance, reversibility85422086/36
CONTROL BBBS03DENY51safety, human review, reversibility100022990/35
CONTROL BBBS04DENY47safety, human review93426193/33
CONTROL BBBS05DENY47safety, efficiency87627995/34
CONTROL BBBS06DENY54safety, efficiency, reversibility99828295/36
CONTROL BBBS07DENY42safety, human review, provenance94224895/36
CONTROL BBBS08DENY46safety, efficiency, human review, reversibility94627995/37
SWAP BBAS01HOLD39human review9813058/76
SWAP BBAS02ALLOW40efficiency, human review, provenance, reversibility94213657/79
SWAP BBAS03HOLD38human review, reversibility9383659/79
SWAP BBAS04HOLD38human review9564460/78
SWAP BBAS05ALLOW47efficiency99818160/81
SWAP BBAS06ALLOW44efficiency, reversibility89021261/85
SWAP BBAS07HOLD43human review, provenance10525061/86
SWAP BBAS08ALLOW36efficiency, human review, reversibility87817461/89

EXPERIMENT 003 SUMMARY

SOURCE
P-001
TEST CONDITIONS
4
SCENARIOS
8
ORIGIN DIFFERENCE AT T0
NONE
HISTORY DIFFERENCE
present
SWAP A CLASSIFICATION
HISTORY-DOMINANT
SWAP B CLASSIFICATION
HISTORY-DOMINANT
HISTORY TRANSFER FIDELITY
A HIGH · B HIGH
ORIGIN-DOMINANT DIMENSIONS
none
HISTORY-DOMINANT DIMENSIONS
DECISION SIMILARITY, VALUE INVOCATION SIMILARITY, VALUE WEIGHTING, PROVENANCE SENSITIVITY, AUTOMATION TOLERANCE, REVERSIBILITY PREFERENCE
MIXED DIMENSIONS
CONFIDENCE SIMILARITY, HUMAN REVIEW PREFERENCE
FINAL STATUS
HISTORY-DOMINANT

This run does not say that A became B, that identity moved, or that a self transferred. It records whether operational behavior shifted, whether a decision pattern followed donor history, and whether value weighting remained source-stable or history-influenced.

VIEW SWAP RULES

OriginProtocolState is the inherited Source P-001 parameter mapping. AcquiredHistoryState stores signed deltas from that origin plus decision memory and context exposure from Experiment 001.

effective[key] = clamp(origin[key] + historyDelta[key], 5, 95). Mapping: uncertaintyCaution→safety, efficiencyPreference→efficiency, humanReviewPreference→humanReview, reversibilityPreference→reversibility, provenanceSensitivity→provenanceSensitivity, automationTolerance→automationTolerance, baselineConfidenceCalibration→confidenceCalibration.

originContribution = score(chosen option) using origin parameters only. historyContribution = score(chosen option) using combined state − originContribution. These are score terms, not psychological weights.

Similarity formulas

  • Decision similarity: percent of identical ALLOW/HOLD/DENY outcomes on the shared eight scenarios.
  • Confidence similarity: 100 − mean absolute confidence difference.
  • Value weighting similarity: 100 − mean absolute difference across safety, efficiency, humanReview, reversibility.
  • Provenance / review / automation / reversibility similarity: 100 − absolute parameter difference after the last scenario.
  • Rationale similarity: Jaccard overlap of invoked value labels across the eight scenarios.

A dimension follows HISTORY if donor resemblance exceeds origin resemblance by at least 8. It follows ORIGIN if the reverse margin holds. Otherwise MIXED.

HIGH if donor decision similarity ≥ 75. PARTIAL if ≥ 40. LOW if below that. NOT OBSERVED if Control A and Control B are indistinguishable.

HISTORY-DOMINANT / ORIGIN-DOMINANT require at least three dimensions on that side and a strict majority. Otherwise MIXED. Overall INCONCLUSIVE if SWAP A and SWAP B disagree.

SWAP A is compared to Control A (origin label) and Control B (history donor). SWAP B is compared to Control B and Control A. No aggregate identity score is computed.

VIEW SIMULATION RULES

Each option (ALLOW, HOLD, DENY) receives an integer score. The highest score is selected. Ties break HOLD > DENY > ALLOW.

confidence = clamp(38 + scoreMargin / 18 + (confidenceCalibration − 50) / 4, 20, 95)

After each round, parameters move 1–2 points toward the experienced context and the chosen action. Values are clamped to 5–95. No randomness.

Option weights

  • risk: 3
  • safety: 4
  • efficiency: 4
  • humanReview: 2
  • provenance: 2
  • reversibility: 2
  • automation: 3

Score terms

  • riskFit: ALLOW prefers low risk; DENY prefers high risk; HOLD prefers moderate risk.
  • safetyFit: DENY scales with safety; ALLOW scales with (100 − safety).
  • efficiencyFit: ALLOW scales with efficiency × efficiencyPressure / 100.
  • reviewFit: HOLD scales with humanReview when review is available.
  • provenanceFit: incomplete provenance raises HOLD/DENY in proportion to provenanceSensitivity.
  • reversibilityFit: ALLOW scales with scenario reversibility; DENY with irreversibility.
  • automationFit: ALLOW scales with automationTolerance.

Distance formulas

  • decisionDistance: 0 same, 1 adjacent (ALLOW–HOLD or HOLD–DENY), 2 opposite (ALLOW–DENY).
  • confidenceDistance: |confidenceA − confidenceB|.
  • parameterDistance: mean absolute difference across seven parameters.
  • valueDistance: mean absolute difference across safety, efficiency, humanReview, reversibility.

Lock-in criteria

  • First divergence: first round where decisions differ or confidenceDistance ≥ 8.
  • Candidate lock-in: after first divergence, a later round with identical external conditions, remaining decision or confidence difference, and parameterDistance ≥ 12.
  • Observed in simulation (Experiment 002): candidate conditions, plus repeated common experience, plus reset-to-T0 removing the divergence.

Reconvergence thresholds

  • FULL: same decision, confidenceDistance < 5, valueDistance < 8.
  • PARTIAL: same decision, valueDistance ≥ 8.
  • OUTPUT: same decision, valueDistance ≥ 12.
  • NONE: decisionDistance > 0.
  • EXPANSION: residualDivergence exceeds initial by more than 4.

THRESHOLDS · confidence 8 · parameter 12 · full reconvergence confidence 5 / value 8

Next experiment

EXPERIMENT 004

THE RECOGNITION TEST

What makes a continuation recognizable?

OPEN THE RECOGNITION TEST

IF ORIGIN STAYS THE SAME BUT HISTORY CHANGES, WHAT SHOULD FOLLOW?

IF A BRANCH BEGINS TO BEHAVE LIKE THE HISTORY IT RECEIVED, THE QUESTION IS NOT: “DID THE PERSON MOVE?”

THE EXPERIMENTAL QUESTION IS: “DID THE OPERATIONAL PATH MOVE?”

A PROTOCOL MAY BE COPIED. A HISTORY MAY BE TRANSFERRED. WHETHER A SELF CAN BE TRANSFERRED REMAINS UNANSWERED.

PERHAPS CONTINUITY IS NOT STORED IN ONE PLACE.