Minimal-Difference Matrix¶
Comparison artifact — instantiates Minimal-Change Counterfactual Construction
Lays actual against candidate worlds cell by cell so the single, smallest set of differing facts is visible at a glance.
The Minimal-Difference Matrix is a grid that sets the actual world beside one or more candidate counterfactual worlds, fact by fact, so that exactly which cells differ — and how few — becomes visible at a glance. Its defining move is dissection: it exposes the anatomy of a difference, cell by cell, marking each fact as identical (preserved) or differing (modified) across the columns. It records, for every candidate, precisely what was changed and what was held fixed, making the "difference footprint" of each world countable. It does not collapse a world to a single closeness score or rank the candidates into an order — that is the ranking table's job — and it does not sweep a neighborhood for robustness. Its product is a legible map of where the differences between worlds actually sit.
Example¶
A failure-analysis engineer wants to know why one pump unit seized on the line while its sister unit ran flawlessly. She builds a Minimal-Difference Matrix. The rows are the two units — the failed one and the nominal one; the columns are every build and operating parameter she can enumerate: bearing supplier lot, lubricant grade, torque spec, ambient temperature, duty cycle, coolant flow rate, housing batch, and more. Cell by cell she enters each unit's value and marks it identical or differing. Of forty parameters, thirty-eight match to tolerance; two differ — the bearing lot and the coolant flow. The matrix has isolated the minimal difference: the counterfactual "if the failed unit had had the nominal unit's bearing lot and coolant flow" changes only those two cells, and every other fact is demonstrably held fixed because the grid shows it already was. That is what turns a vague "something must have been different" into a precise, minimal, testable edit — and, by construction, an honestly small one.
How it works¶
- Choose the comparison dimensions. Pick columns broad enough to cover every plausible source of difference; a cause left out of the grid cannot be seen.
- Enter actual and candidate values. Fill each cell with the value that world holds for that fact.
- Mark preserved or modified. Flag each cell identical (within a stated tolerance) or differing.
- Read off the difference footprint. The modified cells for a candidate are its set of departures from actual.
- Prefer the smallest spanning footprint. Favor the candidate with the fewest differences that still covers the outcome, and treat the identical cells as the held-fixed set.
Tuning parameters¶
- Dimension coverage — how many parameters the columns span. Miss one and the "minimal" difference is an illusion — the real difference hides in an absent column.
- Cell resolution — exact match versus a tolerance band for calling two values "identical." Loose tolerance can mark a real difference as a match.
- Candidate count — how many worlds sit side by side. More columns reveal patterns but crowd the grid.
- Highlight rule — what magnitude of difference is flagged as a difference worth noting versus ignored as noise.
- Column ordering — arranging dimensions by suspected relevance so the load-bearing differences surface first.
When it helps, and when it misleads¶
Its strength is that it makes a difference footprint explicit and countable, operationalizing Mill's method of difference: when two cases share every circumstance but one and differ in the outcome, that one circumstance is the difference-maker.[n1] The grid is what lets an analyst see that only one or two facts differ, rather than asserting it.
Its failure mode is the hidden confound: if the true difference sits in a column the matrix never listed, a listed difference gets wrongly blamed, and the apparent minimality is fake. A subtler trap is treating "identical within tolerance" as truly identical when a small logged gap is actually doing the work. The signature misuse is column curation — selecting dimensions so that only the convenient difference appears. The guarding discipline is to strive for column completeness, record tolerances explicitly, and treat any unexplained outcome gap as a sign that a relevant column is still missing, so an unlisted variable cannot masquerade as a matched one.
How it implements the components¶
modification_trace_record— the matrix is, cell by cell, the record of exactly which facts each candidate world changed versus held fixed.preservation_set— the identical-marked cells constitute the explicit set of facts held fixed across the comparison, read directly off the grid.
It exhibits the anatomy of a difference but assigns no order. The closeness_ordering_rule that collapses each world to a single score and ranks the candidates belongs to its nearest relative, Closest-World Ranking Table: this matrix dissects one comparison fact by fact, while that table sorts whole worlds by closeness.
Related¶
- Instantiates: Minimal-Change Counterfactual Construction — the Matrix exposes the minimal difference footprint the whole discipline is trying to keep small.
- Consumes: Actual Fact Ledger supplies the actual-world column's values.
- Sibling mechanisms: Closest-World Ranking Table · Nearby-World Sensitivity Review · Counterfactual Preservation Checklist · Gratuitous Change Challenge · Actual Fact Ledger
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The mechanism compares actual and candidate worlds cell by cell, computes each difference footprint, and identifies the smallest footprint that still covers the outcome.
Nearest alternative: Decision, Gate & Allocation — It may recommend a candidate, but that selection is downstream of the defining comparison and minimization analysis.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: Minimal-change counterfactual comparison was explicitly developed in philosophical theories of causation and possible worlds.
Related originating lineages:
- Law & Governance — Legal counterfactual reasoning compares material factual differences between actual and alternative cases.
- Statistics & Experimental Design — Controlled comparison supplies the one-factor-difference ideal.
Review resolution: Both independent reviews place the primary provenance in philosophy. The queued differences (alternate_origin_disagreement, origin_mode_disagreement, domain_reach_disagreement, encyclopedia_synthesis_disagreement) concern secondary metadata, not primary lineage. The final retains law_governance, statistics_experimental_design only where a reviewer supplied a formative-lineage rationale; downstream use or broad applicability by itself is not treated as origin. origin_mode=cross_disciplinary_synthesis because the supplied rationales identify formative contributions that are composed in the mechanism's present form. domain_reach=universal records established application breadth separately from provenance. confidence=medium preserves the more cautious evidence assessment. encyclopedia_synthesis=true records whether either reviewer identified deliberate corpus-level composition.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
[n1] John Stuart Mill's method of difference isolates a cause by comparing two cases alike in every respect but one; the differing circumstance, present with the effect and absent without it, is the cause. The matrix is the tabular form of that comparison. ↩