Writing hypothesis pages

This SOP covers type: hypothesis pages — the synthesis layer that ties wiki evidence to testable claims about aging biology. These pages are how the wiki helps with troubleshooting human aging: when a finding lands, which hypothesis does it update?

Core principle: synthesis MOC, not duplicate truth

Hypothesis pages are synthesis MOCs over atomic facts — they aggregate, weigh, and judge evidence that lives elsewhere on the wiki. They do NOT re-state primary findings.

Concretely:

Each evidence claim is a one-sentence summary + wikilink to the verified atomic page
Source of truth remains the study / protein / intervention / phenotype page
If the atomic page gets updated, the hypothesis page can become stale — flag in lint pass
If a claim has no verified atomic page to point to, tag #gap/unsourced and surface

Anti-pattern: restating a primary finding’s n, p-value, mechanism on the hypothesis page. That duplicates and decays. Instead: “Igf1r+/- mice extend lifespan ~26% combined sexes — see insulin-igf1 verified.” One sentence, one wikilink.

Two treatment modes

Different hypotheses warrant different formats. Choose at seeding time and document the choice in the frontmatter.

Mode A: evidence-aggregating

Use when the hypothesis makes specific testable predictions and the wiki contains evidence directly bearing on them.

Examples: free-radical-theory-of-aging, hyperfunction-theory, information-theory-of-aging, hypotheses/negligible-senescence.

These pages function as active research dashboards — they should evolve as evidence accumulates.

Mode B: conceptual frame

Use when the hypothesis is a way of thinking about aging that organizes broader phenomena rather than making sharp testable predictions.

Examples: disposable-soma-theory, antagonistic-pleiotropy.

These pages are lighter — they explain the framework and point to where specific predictions get tested elsewhere.

Structure: Mode A (evidence-aggregating)

---
type: hypothesis
aliases: [...]
proposed-by: [Author Name]
proposed-year: YYYY
status: active | contested | superseded | confirmed | falsified
treatment-mode: evidence-aggregating
key-evidence-for: ["[[studies/...]]", "[[interventions/...]]"]
key-evidence-against: ["[[studies/...]]"]
related-hallmarks: ["[[hallmark-name]]"]
verified: false
---
 
> ⚠️ Auto-extracted [date] — not yet verified. Synthesis pages can become stale as atomic pages update; flag for lint pass review.
 
# Hypothesis name
 
## The claim
 
[1-2 sentences stating exactly what the hypothesis asserts. Be precise — vague hypotheses can't be tested.]
 
## Status: [active | contested | superseded | confirmed | falsified]
 
[1-2 sentences on the current standing and why. Cite the most decisive evidence.]
 
## Key predictions
 
The hypothesis predicts:
 
1. [Prediction A — should be falsifiable]
2. [Prediction B]
3. [Prediction C]
 
## Evidence supporting
 
- **[Specific finding]** — see [[verified-atomic-page]]. (1 sentence summarizing what's in the atomic page; link does the rest.)
- **[Another finding]** — see [[other-verified-page]].
 
## Evidence against
 
- **[Contradictory finding]** — see [[verified-page]]. The atomic page documents the n, p-value, methodology.
- **[Failed-prediction trial]** — see [[intervention-page]] (verified-partial — null result).
 
## What would update this hypothesis
 
[Forward-looking section. What experiments / findings / cohort studies would shift the status? This is what makes the page useful for someone troubleshooting human aging — they know what to look for.]
 
## Related hypotheses
 
- [[other-hypothesis]] — adjacent / competing / complementary
 
## Related hallmarks
 
- [[hallmark-name]] — direct mechanistic link
 
## Related interventions
 
- [[intervention-page]] — directly tests prediction X
 
## Related model organism findings
 
[Where worm/fly/mouse studies sit in the evidence table — distinguish from human evidence.]
 
## Notes / open questions
 
[Synthesis-level commentary. Acknowledge contested data, methodological caveats. This is the "judgment" content; everything else is pointers.]

Structure: Mode B (conceptual frame)

---
type: hypothesis
aliases: [...]
proposed-by: [Author Name]
proposed-year: YYYY
status: active-frame | superseded-frame  # frames don't get falsified the same way
treatment-mode: conceptual-frame
related-hallmarks: ["[[hallmark-name]]"]
verified: false
---
 
> ⚠️ Auto-extracted [date] — not yet verified.
 
# Framework name
 
## The frame
 
[1-2 sentences: how does this framework organize aging biology?]
 
## What it explains well
 
- [Phenomena that fit cleanly into this framework]
 
## What it doesn't easily address
 
- [Phenomena where the framework strains or fails]
 
## Where specific predictions get tested
 
[Pointer to evidence-aggregating hypothesis pages or intervention/protein pages where derived testable claims live.]
 
## Related frameworks / hypotheses
 
- [[other-framework]]
 
## Notes / why this frame matters
 
[Brief synthesis commentary.]

Discipline rules

Every claim must either wikilink to a verified atomic page or carry #gap/unsourced. No restating primary findings without pointer.
Status assignment is a judgment call — document the reasoning briefly in the body. “Falsified” should mean specific predictions failed in well-powered trials, not just “out of fashion.”
Differentiate evidence types:
- Direct test of prediction (highest weight)
- Indirect support (lower weight)
- Mechanism consistent (lowest weight — many hypotheses are consistent with the same mechanism)
Flag immature hypotheses honestly. If most evidence is preclinical, say so. If a hypothesis is exciting but young (e.g., information theory + Yamanaka reprogramming), don’t oversell.
Distinguish “has been falsified” from “predictions never tested.” A hypothesis with no derived testable claims is not “active” — it’s underspecified. Note this if relevant.
Update on atomic-page changes. When a protein/intervention/study page gets a verifier correction, hypothesis pages that cite it may need a refresh. Surface in lint pass.
Never write “Recent studies show…” — link to the specific verified page or omit the claim.

Common failure modes to avoid

Synthesis-without-citation creep — gradual addition of unsourced commentary. Each paragraph should either link to verified pages or be flagged.
Stale evidence lists — atomic-page corrections don’t propagate. Mitigation: lint pass should diff hypothesis pages against changed atomic pages.
False symmetry — listing equal “for” and “against” evidence when the actual standing is asymmetric. The status field forces honesty here.
Rebranding ineligibility as alive — when a hypothesis can’t be tested in humans, that’s not the same as it being “active.” Note “untested in humans” explicitly.

Lint pass checks specific to hypothesis pages

Every wikilink in key-evidence-for / key-evidence-against should resolve to an existing page
Every linked page should have verified: true (or verified-partial with note)
The status field should match the body discussion
#gap/unsourced claims should be tracked in gaps/README.md

🧬 Aging Wiki

Explorer

writing-hypothesis-pages