Convergence–Divergence Rubric¶
Test or assessment — instantiates Independent Evidence Triangulation
A precommitted rating scale that classifies how a set of streams relate — full agreement, partial compatibility, material conflict, or unresolved divergence — before the favored result is known.
The word "converge" does a lot of dishonest work. After results are in, almost any collection of streams can be narrated as "broadly consistent" if you squint, or as "in conflict" if you would rather not act. The Convergence–Divergence Rubric removes that discretion by precommitting — before anyone sees the results — to a fixed scale that says what each level of agreement looks like and what it licenses. Its defining trait is that it is an instrument fixed in advance: it does not investigate a disagreement or price the final confidence; it renders a rating — agreement, partial compatibility, material conflict, unresolved divergence — against criteria written when no one yet knew which answer they wanted.
Example¶
A paleoclimate group is reconstructing summer temperatures for a region over the last millennium, drawing on three proxy streams: tree rings, lake-sediment layers, and ice-core isotopes. Before extracting the final series, they write a rubric. Convergence requires that all three proxies agree on direction and fall within a stated envelope of each other across most of the record. Partial compatibility covers same-direction trends that drift outside the envelope. Material conflict is reserved for opposite-direction signals in the same interval. Unresolved divergence is any conflict that survives the group's comparability adjustments.
When the reconstructions come in, two proxies agree closely, but the ice cores show an opposite swing during one volcanic century. Because the rubric was fixed in advance, the group cannot wave this away as "one noisy stream" — it scores as material conflict for that interval, which under the precommitted stopping rule means the millennium-long claim can be published at high confidence except for that century, which is flagged as insufficiently resolved and handed off for a targeted study. The rubric did not tell them why the ice cores disagreed; it told them, without post-hoc wiggle room, that they were not allowed to call it convergence.
How it works¶
- Write the scale before the data. Each level — agreement, partial compatibility, material conflict, unresolved — gets an operational definition tied to the decision's sensitivity, committed while the result is still unknown.
- Define "material" in substantive terms. Exact numeric identity is rarely required; directional-only similarity is usually too weak. The threshold is set to the swing that would actually change the decision.
- Score the set, don't diagnose it. The rubric assigns a level; it deliberately stops short of explaining why streams diverge, handing material conflicts onward.
- Bind each level to a sufficiency consequence — what score permits closure, what score forces more evidence — so the rating is not merely descriptive.
Tuning parameters¶
- Convergence envelope width — how close streams must sit to count as agreeing. Tight envelopes rarely trigger false convergence but often report "partial" for practically-aligned evidence.
- Materiality threshold — the divergence size that escalates. Set to decision sensitivity, not to statistical significance.
- Level count — a coarse three-band scale is fast and legible; a finer scale captures nuance at the cost of rater disagreement.
- Comparability allowance — how much unit/definition harmonization is permitted before scoring, since over-harmonizing manufactures the very agreement the rubric is meant to test.
When it helps, and when it misleads¶
Its strength is that it makes convergence a falsifiable claim: because the bar was set first, the team cannot retro-fit "the evidence broadly agrees" onto a pile that doesn't. A good rubric behaves like an inter-rater instrument — two people applying it to the same streams should land on the same level.[1]
Its failure mode is brittleness and false comfort. A precommitted threshold set at the wrong place can score genuinely compatible streams as "conflict" (or the reverse), and a rubric applied to streams that secretly share a source will happily rate them "convergent" — it measures whether they agree, never whether they had a right to. The classic misuse is running the comparability allowance hot, transforming units until disagreeing streams slide inside the envelope. The guarding discipline is to freeze the rubric before results, log any mid-stream change to it as a challenge, and never let it substitute for the dependency check that decides whether the agreement was independent in the first place.
How it implements the components¶
convergence_and_divergence_rule— it is that rule, made concrete and precommitted: the operational definitions of agreement, partial compatibility, material conflict, and unresolved divergence.stopping_and_sufficiency_rule— each rating carries a fixed consequence for closure, so a "material conflict" score blocks premature synthesis while a clean "agreement" permits it.
It renders a rating but does not act on it: diagnosing why a conflict exists is contradiction_investigation_path, owned by Contradiction Resolution Workshop, and converting the ratings into a final belief is confidence_calibration_statement, owned by Confidence Update Worksheet.
Related¶
- Instantiates: Independent Evidence Triangulation — supplies the precommitted agreement scale the rest of the lifecycle rates against.
- Sibling mechanisms: Contradiction Resolution Workshop · Confidence Update Worksheet · Cross-Source Corroboration Table · Blinded Parallel Analysis
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: A precommitted rating scale that classifies how a set of streams relate — full agreement, partial compatibility, material conflict, or unresolved divergence — before the favored result is known, making its operative form a bounded evaluation of existing evidence or work that produces a finding or disposition.
Independent corroboration: The frozen evidence defines Convergence–Divergence Rubric as 'A precommitted rating scale that classifies how a set of streams relate — full agreement, partial compatibility, material conflict, or unresolved divergence — before the favored result is known', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Evidence-synthesis methodology cohered predeclared scales for grading agreement, compatibility, conflict, and unresolved divergence before results are known.
Related originating lineages:
- Ethnography & Qualitative Methods — Triangulation practice contributes treating disagreement among streams as information rather than averaging it away.
- Philosophy — Epistemology contributes rules for what differing degrees of convergence warrant.
Review resolution: Predeclared convergence grading synthesizes evidence appraisal, qualitative triangulation, and epistemic rules for interpreting conflict and missing evidence.
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.
References¶
[1] Cohen, J. "A Coefficient of Agreement for Nominal Scales". Educational and Psychological Measurement 20(1), 37–46 (1960). Supports testing whether independent judges classify the same material into the same categories; the specific rubric analogy is an application beyond the paper. registry ↩