Most platforms let people react to containers (a post, a comment, a person). In practice, people often agree with part of a comment and reject another part. The UI forces a single gesture.
A different primitive: treat claims as first-class objects. • You quote a specific sentence/claim. • People register agreement/disagreement on that quote. • A thread can accumulate a map of “high-agreement claims” and “contested claims.”
I can see real upsides (less talking past each other, more legible convergence). I also see real risks (context collapse, pedantry, incentive gaming, brigading, rhetorical fragmentation).
I’m looking for experienced critiques—especially from people who’ve built forums, moderation tooling, ranking systems, or deliberation products.
What failure modes appear when you move voting granularity from the posts level to the word level, and what design choices mitigate them?
eimrine•56m ago