When this rubric applies
The argument’s conclusion makes an empirical claim about how the world is, was, or will be — "X causes Y," "the rate of Y has fallen," "intervention I will reduce Y." The disagreement turns on the strength of evidence, not on values.
Not appropriate when: the disagreement is about what follows from a finding (use POLICY-TRADEOFF), about a question of legal authority (use STATUTORY-INTERP), or where the comparison is structural to a past episode (use HISTORICAL-ANALOGY).
Criteria
1. Quality of evidence cited
weight 25%
Are sources primary, peer-reviewed, and appropriate to the claim’s scope?
10Cites peer-reviewed primary studies, named, with the right design for the claim (RCT for causal, cohort for correlational, etc.).
7Cites credible secondary sources or single primary studies; minor mismatch between study design and claim type.
4Relies on press summaries or single anecdotes; the cited evidence does not actually support the specific claim.
1Cites no evidence, or cites discredited or retracted work.
2. Treatment of confounders & alternatives
weight 20%
Does the argument acknowledge the obvious alternative explanations and rule them in or out?
10Names the leading alternative explanations and engages each — rules out, controls for, or concedes uncertainty.
7Mentions one alternative; ignores another that’s clearly relevant.
4Treats the cited correlation as if causation were obvious; no engagement with confounders.
1Reverses the direction of causation, or cherry-picks the favorable interpretation without acknowledgment.
3. Calibration to the actual literature
weight 20%
Does the argument’s confidence match the field’s actual state of knowledge?
10Acknowledges where the evidence is mixed; quantifies effect sizes; cites meta-analyses where they exist.
7Largely calibrated; one place where confidence outruns the cited evidence.
4States contested findings as settled; or states established findings as fringe.
1Treats one preferred study as dispositive against a literature that disagrees.
4. Logical structure and coherence
weight 20%
Do the premises actually support the conclusion, or is there a missing inferential step?
10Premises chain cleanly from evidence to conclusion; assumptions are surfaced.
7One unstated assumption does material work but is plausible.
4An inferential gap (e.g., conflating "at the population level" with "for any individual") carries the argument.
1Conclusion does not follow even if all premises were granted.
5. Distinguishes claim from policy implication
weight 15%
When the empirical claim is contested, does the argument keep "is X true?" separate from "what should we do about X?"
10Empirical claim and policy implication are clearly separated; the argument scores even if the policy view is rejected.
7Mostly separated; one place where policy framing colors the empirical assessment.
4The empirical and policy claims are bundled; rejecting one rejects the other.
1Asserts an empirical claim solely because it would be convenient for the preferred policy.
Version history
v4April 2026 · added criterion 5 (claim vs. policy implication) after the microplastics thread surfaced bundling issues
v3January 2026 · sharpened scoring anchors on criterion 2 to penalize unaddressed confounders harder
v2August 2025 · reweighted to bring "calibration to the literature" up to 20%
v1March 2025 · initial publication