It might be good to explicitly state in the hover text over the upvote and downvote buttons that they mean “would like to see more of this” and “would like to see less of this”, rather than the mysterious and vague “like” and “dislike”.
More radically, instead of vague “agree” and “disagree”, one could imagine placing a small probability distribution in each comment and votes consist of marking how much credence you have in whatever that comment is saying. This is more confusing if the comment makes multiple claims, though, but that’s a failure mode of the agree and disagree also.
Perhaps it should be possible to highlight sections of a comment and mark them with probability distributions that pop up when you hover over them and which also subtly color the highlight (divide probabilities into three ranges: red=0-33%, green=33-67%, blue=67-100%, then weight the RGB values by the number of votes in each range), as well as putting a small unobtrusive icon shaped like the probability distribution (perhaps in the margin?) when not hovering...
I just made a bunch of claims all at once… that is indeed a failure mode of this system which is going to regularly occur.
This is more confusing if the comment makes multiple claims, though, but that’s a failure mode of the agree and disagree also.
It’s also more confusing if the original comment made a claim like “The sky is blue, with 70% probability.” Then if a user assigns 40% probability to that comment, it’s not clear whether they mean:
I think it’s 40% likely that the sky is blue.
I think it’s 40% likely that you assign 70% probability to the sky being blue. (E.g., maybe you’re going back and forth about what your true belief is, and I want to weigh in on what I think your view is.)
I think it’s 40% likely that you’re correct in assigning 70% probability to the sky being blue. (E.g., maybe I think you’re underconfident and the true probability is 90%; or maybe I think you’re overconfident and the true probability is 50%; etc.)
I just made a bunch of claims all at once… that is indeed a failure mode of this system which is going to regularly occur.
I think the current system isn’t ideal, but I don’t particularly mind this specific issue. It’s already a problem for upvotes/downvotes, and I think upvotes/downvotes are a good feature on net in spite of this. (And it’s at least plausible to me that adding more UI complexity in order to let someone upvote/downvote parts of posts/comments would be net-negative.)
Part of why I’m fine with this issue is that I think it’s just good for people to be separately tracking agree/disagree and good/bad. Even if they don’t end up voting ‘agree/disagree’ that often, I expect positive affects from the mental activity alone. (E.g., prompting people to think in this mode might cause them to notice that they agree with the first half of a comment but not the second half; in which case we’re already making good things happen, whether or not they write a follow-up comment explicitly saying ‘I agree with the first half but not the second’.)
It might be good to explicitly state in the hover text over the upvote and downvote buttons that they mean “would like to see more of this” and “would like to see less of this”, rather than the mysterious and vague “like” and “dislike”.
More radically, instead of vague “agree” and “disagree”, one could imagine placing a small probability distribution in each comment and votes consist of marking how much credence you have in whatever that comment is saying. This is more confusing if the comment makes multiple claims, though, but that’s a failure mode of the agree and disagree also.
Perhaps it should be possible to highlight sections of a comment and mark them with probability distributions that pop up when you hover over them and which also subtly color the highlight (divide probabilities into three ranges: red=0-33%, green=33-67%, blue=67-100%, then weight the RGB values by the number of votes in each range), as well as putting a small unobtrusive icon shaped like the probability distribution (perhaps in the margin?) when not hovering...
I just made a bunch of claims all at once… that is indeed a failure mode of this system which is going to regularly occur.
It’s also more confusing if the original comment made a claim like “The sky is blue, with 70% probability.” Then if a user assigns 40% probability to that comment, it’s not clear whether they mean:
I think it’s 40% likely that the sky is blue.
I think it’s 40% likely that you assign 70% probability to the sky being blue. (E.g., maybe you’re going back and forth about what your true belief is, and I want to weigh in on what I think your view is.)
I think it’s 40% likely that you’re correct in assigning 70% probability to the sky being blue. (E.g., maybe I think you’re underconfident and the true probability is 90%; or maybe I think you’re overconfident and the true probability is 50%; etc.)
I think the current system isn’t ideal, but I don’t particularly mind this specific issue. It’s already a problem for upvotes/downvotes, and I think upvotes/downvotes are a good feature on net in spite of this. (And it’s at least plausible to me that adding more UI complexity in order to let someone upvote/downvote parts of posts/comments would be net-negative.)
Part of why I’m fine with this issue is that I think it’s just good for people to be separately tracking agree/disagree and good/bad. Even if they don’t end up voting ‘agree/disagree’ that often, I expect positive affects from the mental activity alone. (E.g., prompting people to think in this mode might cause them to notice that they agree with the first half of a comment but not the second half; in which case we’re already making good things happen, whether or not they write a follow-up comment explicitly saying ‘I agree with the first half but not the second’.)