Admins cannot determine the motive, which is crucial for distinguishing the ‘vandalism’ case from the ‘I think everything this user writes is bad and I want less of it’. The latter is aggressive but a necessary part of voting. Some users are simply lower quality than others. This should not be construed as a verdict as to the quality of user:ialdabaoth’s contributions.
Given the described voting patterns, I think the exact motivation in this case and the other cases is a) easy to guess at and b) not that relevant for actually how it should be dealt with. Suppose for example that someone is doing this who really liked iadobaoth’s views and is doing this as part of a convoluted scheme to generate more sympathy for him. That would still be unacceptable.
What admins (or for that matter automated software) could do is to publish the information. This can be done without determining motives. I’m thinking, e.g., that if user A mass-downvotes user B (according to some heuristic whose details are important but I’m not going to try to nail down here) then (1) when you go to the user page of A or B, that information is listed near their karma score, and maybe (2) when you hover over user B’s karma score on any page, it indicates the fact. (“Mass-downvoted by Dias, gjm, and 6 others” or something.)
In that case, it would still be possible to downvote everything a user posts, if you truly think everything they’ve done is bad. You’d just have to accept that in that case your decision to do so would be publicly visible. In cases where someone really just is posting a lot of stupid stuff, this would probably enhance your reputation among sensible LW users, rather than hurting it.
(Roughly what might that heuristic look like? Something like this: “A has mass-downvoted B if A has downvoted at least 20 of B’s contributions, has downvoted at least twice as many of B’s contributions as s/he’s upvoted, and there’s some sequence of at least 20 consecutive contributions from B of which A has downvoted at least half.” Every single detail of that is nitpickable; I’m just gesturing vaguely towards the sort of thing that might work well.)
In principle, yes. In practice, I bet most pairs of users don’t occur (“long-tail” users neither vote much nor get voted on much). And the software needs to keep track of every vote anyway, so it can show you what votes you’ve already made and stop you voting twice.
Yes, you are right. Actually, if the underlying database associates each vote with a post already, you don’t need any additional data structures, you can do it all through SQL queries...
Admins cannot determine the motive, which is crucial for distinguishing the ‘vandalism’ case from the ‘I think everything this user writes is bad and I want less of it’. The latter is aggressive but a necessary part of voting. Some users are simply lower quality than others. This should not be construed as a verdict as to the quality of user:ialdabaoth’s contributions.
Given the described voting patterns, I think the exact motivation in this case and the other cases is a) easy to guess at and b) not that relevant for actually how it should be dealt with. Suppose for example that someone is doing this who really liked iadobaoth’s views and is doing this as part of a convoluted scheme to generate more sympathy for him. That would still be unacceptable.
What admins (or for that matter automated software) could do is to publish the information. This can be done without determining motives. I’m thinking, e.g., that if user A mass-downvotes user B (according to some heuristic whose details are important but I’m not going to try to nail down here) then (1) when you go to the user page of A or B, that information is listed near their karma score, and maybe (2) when you hover over user B’s karma score on any page, it indicates the fact. (“Mass-downvoted by Dias, gjm, and 6 others” or something.)
In that case, it would still be possible to downvote everything a user posts, if you truly think everything they’ve done is bad. You’d just have to accept that in that case your decision to do so would be publicly visible. In cases where someone really just is posting a lot of stupid stuff, this would probably enhance your reputation among sensible LW users, rather than hurting it.
(Roughly what might that heuristic look like? Something like this: “A has mass-downvoted B if A has downvoted at least 20 of B’s contributions, has downvoted at least twice as many of B’s contributions as s/he’s upvoted, and there’s some sequence of at least 20 consecutive contributions from B of which A has downvoted at least half.” Every single detail of that is nitpickable; I’m just gesturing vaguely towards the sort of thing that might work well.)
That might actually be a devil-in-the-details thing, and gameable by rules mechanics, too.
Another point is technical issues—you want to keep track of interactions (up/downvoting) between pairs of users and that’s an O(n^2) problem.
In principle, yes. In practice, I bet most pairs of users don’t occur (“long-tail” users neither vote much nor get voted on much). And the software needs to keep track of every vote anyway, so it can show you what votes you’ve already made and stop you voting twice.
Yes, you are right. Actually, if the underlying database associates each vote with a post already, you don’t need any additional data structures, you can do it all through SQL queries...