The encyclopedia that edits back

Definition

Mirror hypothesis. Wikipedia merely summarizes published science. Influence flows one way, from journals to encyclopedia, and never back.

Seed hypothesis. Wikipedia also shapes science itself: what researchers notice, cite, and even how they phrase their findings.

Randomized controlled trial (RCT). An experiment that randomly assigns a treatment to some subjects and withholds it from comparable controls, so outcome differences can be attributed to the treatment rather than to pre-existing differences between the groups.

The respectable objection

Fairness first. The mirror hypothesis is not a strawman. Wikipedia's core policies prohibit original research and require verifiable citations to already-published sources. On paper, the encyclopedia is a lagging indicator by construction: someone publishes, someone else summarizes. Anyone who has watched editors argue over citation formatting may find the seed hypothesis difficult to take seriously. The obvious objection is a good one.

It is also, apparently, wrong in one direction. A paper by Neil Thompson and Douglas Hanley tested the claim directly with an RCT, and the working paper has since drawn well over a hundred citations Science is Shaped by Wikipedia.

The experiment

The design was refreshingly blunt. The researchers found scholarly chemistry papers that deserved encyclopedia coverage but had none, wrote short entries citing those source papers, and randomized which entries actually went live during the study window. Comparable topics stayed unwritten until the study ended. Then they watched the literature.

Context matters for why this is plausible at all: Wikipedia is one of the most accessed websites in the world, including by scientists the trial paper. As one write-up put it, "Wikipedia is everyone's go-to source. Even scientists." ScienceAlert.

The language finding

The citation effect could, in principle, be waved away as attention routing: the entry made a paper easier to stumble upon. The phrasing effect is harder to dismiss. The experiment specifically traced how the encyclopedia influences research write-ups, finding Wikipedia's wording bleeding into papers Nature's 2017 coverage. When researchers borrow sentence-level language, the encyclopedia is functioning as a drafting input, not just an index. Deadpan summary: a summary source that changes the thing it summarizes is no longer only a summary.

Limits

• One field. Chemistry, with data from the mid-2010s. Extrapolating to high-energy physics or clinical medicine is a guess dressed as a conclusion.
• Stubs, not features. Entries of about a hundred words are minimal treatments. Full articles might produce larger or qualitatively different effects.
• Citations are proxies. The trial measured attention, not correctness. More citations is not automatically better science.
• The platform moves. Editing norms, traffic patterns, and machine reuse of Wikipedia text have all shifted since the trial ran, so the magnitudes may not travel.

These are real caveats, and none of them rescues the pure mirror hypothesis. The direction of the causal arrow was the question, and the experiment answered it. The defensible position is the boring one: Wikipedia is both mirror and supply line, and the mirror part was never in doubt. What changed is that the supply line now has experimental evidence behind it.

Sources

• Science is Shaped by Wikipedia: Evidence From a Randomized Controlled Trial (SSRN)
• Wikipedia shapes language in science papers (Nature, 2017)
• Wikipedia has become a science reference source even for scientists (ScienceAlert, 2018)