Starting this weekend, LessWrong will be displaying karma notifications in the top-right corner, telling you about when you've been upvoted or downvoted. You can click the star icon to see which of your posts and comments have been voted on, and how much their score has changed.
This works a little differently from how most web sites do it. I've noticed a tendency, in myself and others, to sometimes obsessively refresh pages hoping to have gotten Likes. The rest of the LessWrong team has noticed this too. That's why on LessWrong, by default, this won't work; karma-change notifications are grouped into daily batches, so after you've checked it, you won't be notified of any additional votes until the next day.
While I hope that this prevents people from using LessWrong in ways that they don't endorse, daily batching might not be a strong enough safeguard for everyone. If you find yourself tempted to check LessWrong more often than you think is ideal, you can change the karma-notifier batches to weekly, or disable them entirely. On the other hand, if checking and posting on LessWrong is something you'd rather do more of (because you'd otherwise do something less valuable), you can set it to real-time.
As a nonprofit with no advertising, LessWrong is under a different set of incentives than most websites. We worry a lot about the trade-off between engagement and addictiveness. I want people to use the site in ways they reflectively endorse, even if that means using it less. I think LessWrong should leave that under user control, as much as feasible.
(As a reminder: Voting on LessWrong is a way of saying "I want to see more/less things like this". Upvotes are a way to tell people that their post or comment was worth their time to write, and worth your time to read. It does not necessarily indicate agreement; sometimes a good write-up of an ultimately incorrect idea is what a conversation needs. Conversely, downvotes do not necessarily indicate disagreement; sometimes a correct points is written in a way that's confusing, inflammatory or otherwise detracts from the conversation.)
As always, bug reports are welcome here, via the Intercom widget, or on the GitHub issue tracker.
I was particularly concerned about this feature and ended up pushing very hard for implementing batching and making it easy to disable before we push anything like this.
Zvi in particular has warned in the past about making karma more prominent on LessWrong, and I roughly share his worries. A bunch of initial designs of this had a display of your total karma right next to your user, which I really didn't like, because I think that would push too much in the direction of karma being a direct measure of your status, and would lead to too much social comparison and goodharting. I want new users who show up on the site to feel rewarded when they engage with content, not to feel repeatedly reminded of their relative inferiority to the senior members of the site in terms of karma. (We also decided against displaying total karma in the new user hover-preview which shows you the data a user joined, as well as the number of comments and posts they created, but very intentionally not their total karma)
I do think that giving users full control over the relevant UI elements makes a large difference here, and deals with a good chunk, though not all, of the problems with karma notifications. Batching I think also helps with a good chunk of them, by giving you a lot more granular control over how you want your behavior to be reinforced.
The things that I am most worried about going forward are:
The reversal test is with respect to the norm, not with respect to ways of handling a fixed norm. So imagine that the norm is the opposite, and see what will happen. People will invent weird things like gaging popularity based on number of downvotes, or sum of absolute values of upvotes and downvotes, when there are not enough downvotes. This will work about as well as what happens with the present norm. In that context, the option of "only upvotes" looks funny and pointless, but we can see that it actually isn't, because we can look from the point of view