What breaks when you vote on specific claims instead of whole posts?

[flyblackbox]

For a software project I’m working on I’ve been studying a pattern in online discussions that feel thoughtful yet inconclusive.

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?

[baubino]

I like the original premise for the upvote, which was that you don’t vote on whether you agree but on whether the comment/post is a good contribution to the discussion. The upvote should be a way of saying “good point (even if I disagree)” and the downvote should be saying “this is irrelevant or otherwise doesn’t contribute to the conversation.”

Of course, almost nobody uses the vote this way anymore. While your granular voting is quite interesting in principle, in practicality it seems it would negatively compound the existing problems with the vote system, namely that instead of voting to support the continuation of good faith discussion, everyone is voting to support just their own ideas. That in turn leads to fractious discussion (if we can even call it discussion) where the most popular and well-known ideas are strongly upvoted and continue to circulate, and anything deviating is barely seen. Then you don’t really have a discussion; you just have a series of highly upvoted statements. (See, for example, Reddit.)

[flyblackbox]

This is a strong critique. If voting inevitably drifts toward factional reinforcement, is there any interface you’ve seen that resists that drift? I’m curious to know which product or feature you think solves this the best.

[baubino]

I like that HN doesn’t give voting power to new accounts. And because it takes a while to build the points (or whatever they’re called here) to get the voting power, that time allows newcomers to see and get used to how discussions works here.

I like that Discourse only has an upvote button and no downvote, but that the replies still stay in chronological order. That way there’s an actual flow to the conversation but you can still see which ideas people value.

[compressedgas]

Reminds me of Zest: http://zesty.ca/zest/

[flyblackbox]

Thanks for sharing that. I hadn’t heard of Zest before and it mirrors some of the user interface designs I’ve been working on.

Did you use Zest? What aspects of it worked well or failed in practice? It’s cool that some of the work ended up in Google Wave. I always thought that was an amazing product, I was so sure it was going to replace email and then it ended so abruptly.

[eimrine]

Not every claim is legal to be said. If your system works that good it inevitably becomes related to politics.