Community voting ranks contributions by aggregated member preference, usually upvotes minus downvotes, often weighted by recency. It surfaces broad agreement efficiently but systematically favours familiar ideas, early submissions and confident presentation. A vote count measures how many people approved of something, not whether it was good.
They measure approval, which correlates with quality loosely and unreliably. Voting favours familiar, confidently presented and early-posted material, so a low-scoring item may simply have been unusual or late rather than worse.
Because visibility and votes reinforce each other, and because familiar ideas are quicker to evaluate. Both effects push a community towards its own average over time unless something deliberately counteracts them.
They add information but also add a way to suppress unpopular positions rather than weak ones. Systems with downvotes tend to converge on consensus faster, which is useful for filtering noise and costly for anything genuinely contrarian.
There is no clean threshold, but with fewer than a few dozen votes an outcome mostly reflects who happened to be present. Treat early counts as directional at best, and be suspicious of anyone presenting small numbers as a mandate.