How to make the Internet a nicer place to live

I don't think you can eliminate the ugliness. That's okay. I think if you can somehow cluster the ugly comments together, let the ugly people have that discussion, hidden from view, everything might work out. Perhaps textual analysis is enough, grouping comments by words, clustering by k-means. Similar to how dead accounts can still post here, ugly comments should be accepted, clustered into the ugly pile, and shown only to posters who use similar .. language. If my entire post is 'what a bunch of f__ing sh__heads', it wouldn't be hard to cluster. Then the public/non-logged-in view would default to the 'civil' cluster, with showdead/showugly as options, if you're into that kind of thing.

I don't think you should stick ratings on people, just to comments. If people have any input, Zed Shaw would have a really poor rating, let alone Lennart Poettering. There's nothing inherently wrong with their comments, of course, so they'd likely be clustered 'civil'.

I'm not too familiar with the domain-relevant quiz: that intrigues me. Could you imagine, a did you really read the article captcha? Of course the /. editors can barely fix the typos; creating reading comprehension quizzes is an enormous task. But if I'm the author of a post, I might could do it.

Some of the online game matching services are starting to do this - I think both League of Legends and DOTA group people based on the number of "reports" they've had against players.

Couple of thoughts, unsorted brain dump:

* Ranking/Karma systems can and will be abused. See the digg bury brigade, etc. It's really difficult to separate honest feedback from feedback deliberately tailored to vanish unpopular opinion, or hurt people. (It's not like eBay ratings aren't gamed, either. Or Amazon ratings/reviews)

* Related to that: One person's troll is another person's idol. Post on any political topic if you don't believe that :)

* Related: I believe comment-based rating works slightly better than user-based rating. Just because it takes a lot of effort to silence a user by downvoting every single comment. It still allows burying unpopular opinion

* I believe rating systems are gamed because rating is "free". If you only had limited rating capital to expend, I'd assume most people would not want to waste it on burying things.

* On the flip side, reputation doesn't matter to many people because it is assigned to online IDs, not actual people. I.e. reputation is free, as well.

* Instead of a narrow AI, I'm wondering if a simple Bayesian filter might do the trick for the vast majority of Internet trollery. Yes, it's gameable, but as troll filters become more sophisticated over time, gaming becomes OK. [3]

* An immutable ledger of past reputation transactions might be meaningful. (Yes, I'm on the Bitcoin bandwagon with that, sorry). It does not solve the issue of multiple accounts with different reputations, but it's in each participants interest to have each individual entity represent consistent reputation.

* I believe any reputation system will have a social component. What your friends think about somebody you never met matters more than what random strangers think, because chances are your friends and you have at least similar value systems.

* Society-level interventions might start with educating people in "direct communication" [1] and "nonviolent communication"[2]. And yes, education in general helps too. I'm not sure "biological interventions" are without ethical issues.

* You do increase the risk of an information bubble - people might confuse civility with agreement. But I do think that over time, that might sort itself out. In my experience, people I consider of high integrity tend to rank other people's integrity based on behavior, not opinion or political leaning. I might be an optimist.

* It seems I use politeness and trust interchangeable. I'll have to ponder that some more, but it seems fundamentally correct to me. I care about respectful politeness, not fake politeness with ensuing backstabbing.

[1] https://csde.washington.edu/~mbw/direct-and-indirect-communi...

[2] http://en.wikipedia.org/wiki/Nonviolent_Communication

[3] http://xkcd.com/810/

I think another aspect of improving the problem of ugly comments is figuring out more ways to encourage insightful, constructive comments while making it worthless to post hateful things. HN seems to have figured out how to do this reasonably well, but it's somewhat specific to the community that has grown up around it and that community has shown a willingness to defend itself. How can a similar system work for more general audiences at news sites and forums, much less sites like YouTube where the comments are almost universally horrible?

I think making incivility even worse than worthless would be a start. Make it a waste of the poster's time. Make it so the ugliness can be completely ignored, taking away the influence and attention that the poster seeks. Make it reflect badly on the poster somehow so they will have an incentive to not post their hate.

I always thought these kinds of policies would have to come from the community, but the idea of AI doing some of the work is intriguing. It may seem far-fetched now, but like so many things, starting the work and making progress will probably lead to great advances.

Maybe things like Discourse will make progress here, and I know Jeff Atwood believes just as passionately about civility on the internet as Eric Sink expresses in his post. But those kinds of systems would have to be adopted more universally to make that big of a difference. I sincerely hope that happens.

I also like ericsink's solution, and am also interested in system-level approaches.

One thing I noticed is that people are nastier when broadcasting to a large audience than when talking to their opponent in private. Maybe we can make nastiness less rewarding by making a forum that's somehow based on private one-on-one discussions, which get opened up later?