Norms must be observable, not merely stated
Aliases: norm observability · norms by demonstration · learning from exemplars
What it is
Community norms — what content is welcome, what behavior gets removed — reach newcomers mainly through watching what others do, not through reading written rules. Codified norms are abstract and long; even a newcomer who reads the whole policy page cannot map "maintain a neutral point of view" onto the paragraph they are about to post. Observable norms — precedents, the shape of well-received posts, visible instances of others being reverted — are the textbook newcomers actually use. The design corollary: a community that maintains only a rule page, without curating observable demonstrations, has norms its newcomers will never learn.
Why it happens
Norm acquisition is fundamentally situational matching: knowing the rule and knowing how it applies to the case at hand are different things, and the second can only be induced from concrete cases. Observational learning is therefore the main channel — seeing how similar content gets written, evaluated, and treated carries the "situation → rule" mapping that rule text lacks. Written rules also accrete: a policy page is the sediment of past disputes among veterans, each clause carrying local history the newcomer lacks, so what they read is a set of unexecutable abstractions. The enforcement side depends on observability just as much: when a rejection cites a clause with no precedent and no specific location, the newcomer cannot locate what they did wrong and can only conclude "these people don't like me" — how observable enforcement is decides whether it educates or expels.
Studying it
- Paradigm: two complementary routes. One is process-data comparison — relate newcomers' pre-contribution browsing history (did they view similar content and similar enforcement cases) to the survival rate of their first contribution. The other is a controlled presentation experiment — give one group pure rule text and another the same text plus real positive and negative exemplars, then compare first-contribution compliance and rejection rates.
- Variables: independents — norm presentation format (text only / with exemplars / rejection templates citing precedent), amount of pre-contribution observation; dependents — first-contribution compliance, rejection rate, retry rate after rejection.
- Use in interface research: auditing whether rejection templates, onboarding flows, and help centers actually convey norms or merely restate the clauses.
- Methodological caveat: observation quantity and compliance are self-selected (cautious newcomers look before writing anyway), so causal identification requires randomizing the presentation format. "Compliance" conflates "learned the norm" with "afraid to post" — read it together with contribution volume.
Where it stops holding
Observable norms only transmit what the community consistently does. If veterans enforce unevenly — identical content removed one day and kept the next — what newcomers observe is noise, and what they learn is "it's a lottery," which is a consistency problem, not a presentation problem. The other edge: in sensitive-topic communities, the visible enforcement cases skew toward warnings, so the "norm" newcomers observe is risk-avoidance, and contributions trend conservative — not a failure of transmission but a built-in bias of the channel.
Applying it
- Attach one real positive and one real negative exemplar (linked to actual content and its handling) to every clause of the rule page; let the clause serve only as the exemplar's headline.
- Make every rejection or removal carry three elements — exactly where, which norm, and what correct looks like (link a precedent) — with none omitted.
- Give newcomers a norm-dense observation ground: a curated "typical week" feed or a novice task queue where correct demonstrations appear at high frequency.
- Verification: sample rejected newcomer contributions and check whether rejection reasons cite precedent or a specific location; compare first-contribution survival and retry rates before and after a rejection-template revision.