Rubric — Single Claim Evaluation

For the short repeated assignment that carries most of the practice in this course: find one peptide claim in the wild, and work it.

"In the wild" means a real place claims actually appear — a product page, a video, a podcast segment, a press release, a forum post, a magazine article, a friend's confident sentence. Not a textbook, not Appendix A, and not a claim the instructor supplied. The difficulty of this assignment is almost entirely in the first step, and it is the step the course's other assessments cannot exercise: a claim in the wild does not arrive with a population, an endpoint, or a comparator attached, and extracting an evaluable proposition from a marketing sentence is the skill.

Expect one to two pages. Three to five of these across a term is the right dose; students improve noticeably between the second and the third.

Required format. Every submission ends with the four-line rating, regardless of how long the analysis is:

CLAIM:            [what, in whom, measured how]
EVIDENCE:         [best available, described by design]
RATING:           [✅ / ⚠️ / ❌ / 🔬]  +  [date]  +  [if ❌: absent / present-and-negative]
WOULD CHANGE IF:  [the specific finding that would move it]

The rubric

Criterion Exemplary Proficient Developing Insufficient
1. The source, captured honestly Source identified with enough detail to be located. The claim is quoted or described in its own words, not improved. Where the source is vague, the vagueness is preserved and analyzed rather than resolved on the source's behalf. Source identified and claim fairly represented; light paraphrase that does not change the strength of the claim. Claim paraphrased into something more reasonable or more absurd than the source made it. Source thinly identified. No identifiable source, or a claim the student appears to have constructed and presented as found.
2. Restatement in evaluable form Converts the found claim into a proposition with an explicit population and endpoint, and states what had to be supplied to do so — because the source usually did not specify. Where the source is unfalsifiable as written, that is named as the finding. Population and endpoint stated; supplied assumptions mostly identified. Restatement present but one element missing or silently invented. No restatement, or a rating attached to a molecule rather than to a claim.
3. Evidence located and described by design Best available evidence found and described by design, population, endpoint, comparator, size, duration, result. Human and animal evidence distinguished. What is conspicuously absent is stated. Limits of the student's own search are declared. Design elements largely present; search reasonable; absence noted. Evidence summarized by volume or by conclusion sentence; comparator or population missing. Evidence taken from the source's own summary of itself. Study counts substituted for design.
4. Rating with date and justification Rating traces to criterion 3 and nothing else. Dated. If ❌, tagged absent or present-and-negative. Justification survives the challenge "which line of the evidence produced this?" Rating dated and defensible; justification present but partly general. Rating undated, or justified by mechanism, popularity, or approval status. Rating asserted without justification, or contradicted by the student's own evidence section.
5. Falsification condition Names the finding that would move the rating — with population, endpoint, comparator, and rough scale. Up and down conditions distinguished where both apply. Specific enough to be checkable. Present but generic ("more research"). Absent.
6. Numbers handled correctly Any relative figure is paired with the absolute figure, or the omission is explicitly noted as a finding about the source. Surrogate endpoints identified as surrogates, with a sentence on what accepting one assumes. Both handled correctly with minor slippage. A relative figure repeated without conversion or comment; a surrogate treated as a benefit without comment. Numbers copied from marketing material and reported as findings.
7. Marketing pattern recognition Identifies the specific persuasive moves in play — origin story doing evidentiary work, mechanism substituting for outcome, testimonial framing, an appeal to regulatory status, a citation that does not support the sentence it follows. Names them without letting them downgrade the rating (Rule 4). Two or three patterns identified accurately. Patterns noticed impressionistically ("this seems sketchy"). Rating driven by tone of the source rather than by the state of the evidence.
8. Proportionate conclusion Conclusion matches the evidence in strength and in scope. Says what is not known as clearly as what is. Neither oversells nor performs skepticism. Proportionate, with a small overreach in one direction. Conclusion outruns the analysis, or retreats from a conclusion the analysis supports. A verdict on the molecule, the seller, or the buyer rather than on the claim.

Scoring

Eight criteria, Exemplary 4 / Proficient 3 / Developing 2 / Insufficient 0. Maximum 32. Criteria 2 and 4 may be double-weighted if you want the score to track the course's priorities more sharply.

For a repeated assignment, consider scoring criteria 1–5 only on the first submission and adding 6–8 from the second onward. Students who are still struggling to extract an evaluable claim cannot productively receive feedback on pattern recognition, and eight rows of criticism on a first attempt mostly teaches discouragement.

Common grading edge cases

1. The claim is unfalsifiable as written. "Supports your body's natural recovery processes." There is no population, no endpoint, and nothing that could fail. Students often panic and either invent a specific version or declare the assignment impossible. The correct answer is that the unfalsifiability is the finding, and the strongest submissions say so, then rate the nearest falsifiable neighbor and note explicitly that they supplied it. Grade that path as Exemplary.

2. A well-supported claim from a marketing source. Occasionally a product page states something that is actually true and well evidenced. Students who have learned that "marketing = ❌" produce a strained argument against it. This is Rule 4 — never downgrade with distaste — and it is worth a full-credit example when someone gets it right. A ✅ on a claim found in an advertisement is a sophisticated submission, not a naive one.

3. The student cannot find any evidence at all. Frequently the correct result. Insufficient would be wrong here; the question is whether the search was competent and whether the absence was reported as absence. "I searched and found no human trials testing this" is a finding with a rating attached — ❌, evidence absent — and it is often the most useful submission in the batch. What distinguishes Exemplary from Developing is whether the student declared the limits of their own search.

4. The source cites a real study that does not support its sentence. The highest-value case in the assignment and worth naming as an exemplar when it appears. Look for whether the student checked the population and endpoint of the cited work against the claim being made, rather than treating the presence of a citation as the end of the inquiry.

5. The claim is about the student's own compound. Grade the claim. Keep every comment attached to a criterion number. Do not remark on the choice. See the dossier rubric, edge case 7.

6. Dosing, protocol, or sourcing details reproduced from the source. Students quoting a marketing page will sometimes carry a regimen or a vendor name across in the quotation. Course policy applies to the submission regardless of where the text originated: the claim can be quoted; the protocol and the seller must be elided. Return for revision the first time without penalty, and note that eliding it costs the analysis nothing — the claim is evaluable without it.

7. Excellent analysis, no four-line block. The block is not decoration. It is the format that makes Rule 1 violations impossible to hide, and a student who skipped it has usually skipped it because their claim would not fit. Return for completion rather than grading around it.

8. The student rates the source rather than the claim. A submission that concludes the company is untrustworthy, or that the influencer is being paid, may be entirely correct and has still not done the assignment. Both can be true: a dishonest source can make a well-supported claim, and an honest one can repeat an unsupported claim in good faith. Mark criterion 8 down and say exactly that.

Related: Chapter 5 (the method) · Chapter 6 (hype cycle) · Chapter 37 · Appendix D (reading a trial) · Appendix F (red flags) · Appendix H (twenty worked evaluations — the model for this assignment) · Evidence Dossier Rubric · Participation Rubric · Practical Evaluation Exam