5 ms·
Ask PG: Why not AI in place of (or in addition to) comment voting?
So, in the time I've been here, it's been my understanding that there was a gentlemen's agreement of sorts that upvotes go to comments which "add to the discussion", even if the specific point of view is something the voter disagrees with. Here's the thing though: we shouldn't need to vote in order to determine what adds to the discussion. Certainly some fairly straightforward heuristics like Flesch-Kincaid (http://en.wikipedia.org/wiki/Flesch%E2%80%93Kincaid_readability_test) Grade Level times number of replies (possibly adding in a factor for the F-K Grade Level of the replies to weed out long chains of drivel) should be relatively good at picking the best comments in a thread.
I'm not saying that voting should be eliminated entirely, but maybe it becomes just part of the heuristic. Imagine something like BCS standings, but for comments. Even vote tallies could become smarter, with votes from "trusted" users counting more heavily than those from "noobs" or well known rabble-rousers.
Finally, I think it would be really neat (assuming you like my proposal of a combined heuristic in place of voting) if the algorithm was open source. Take the Google view of "if you don't like the results, don't special case the outcomes, improve the algorithm!" This would have the added benefit of giving the world a first-class comment value sorting algorithm vetted by some of the top talent in the world, and dog-food-ed on a site where they care deeply about the quality of the discussion.
- jcr 16y agoFirst of all, pg, rtm, jl, tlb and the rest of the ycombinator crew are in the midst of interviews and deciding which startups to fund. In other words, they are probably very busy so your suggestion is poorly timed. Secondly, there is a place for feature suggestions and discussion: http://news.ycombinator.com/item?id=363 http://news.ycombinator.com/item?id=363 I feel algorithmically defining the comments adding to a discussion might turn out to be overly optimistic, but none the less, it is still a fascinating idea.
- mechanical_fish 16y agoCertainly some X should be good at Y. What exactly is stopping you from doing an experiment, instead of asking us to imagine the results of an imaginary test of an imaginary algorithm? Scrape up a few weeks' worth of comments (perhaps via RSS, if that works; I wouldn't know) and see what you can do with them. Show us side-by-side article pages with the actual order and the Whizbang Algorithm's redisplayed version.
- IdeaHamster 16y agoNot a bad idea...in fact, I'd wage there's a market for this sort of technology. :)
- namank 16y agoAs cool as that would be, it would also take away from the community aspect of this place. Plus I'd always be trying to beat it
- brudgers 16y ago>"Imagine something like BCS standings" Ok, now that I have spit out the vomit and brushed my teeth... Some new accounts are used as throw-aways and contain great information. Other new accounts are created by intelligent thoughtful people who just discovered HN or who have been lurking and finally feel like they have something to contribute. Rabble rousing often furthers discussion. That's not to say that AI moderating isn't an interesting idea. The BCS is just a really poor example because it's purpose is to generate a made for television event and then dole out the revenue to the major [American] football conferences.