Makes sense. I still think the bar should be a bit lower for criticism, for a couple reasons.
Motivated reasoning means that we’ll look harder for flaws in a critical piece, all else equal. So our estimation of post quality is biased.
Good disagreement is more valuable than good agreement, because it’s more likely to cause valuable updates. But the person writing a post can only give a rough estimate of its quality before posting it. (Dunning-Kruger effect, unknown unknowns, etc.) Intuitively their subconscious will make some kind of “expected social reward” calculation that looks like
Because of human tendencies, social_punishment_for_sloppy_criticism is going to be higher than the corresponding social_punishment_for_sloppy_agreement parameter in the corresponding equation for agreement.
If social_punishment_for_sloppy_criticism is decreased, on, the margin, that will increase the expected values of this calculation, which means that more quality criticism will get through and be posted. LW users will infer these penalties by observing voting behavior on the posts they see, so it makes sense to go a bit easy on sloppy critical posts from a counterfactual perspective. Different users will interpret social reward/punishment differently, with some much more risk-averse than others. My guess is that the most common mechanism by which low expected social reward will manifest itself is procrastination on writing the post… I wouldn’t be surprised if there are a number of high-quality critical pieces of LW that haven’t been written yet because their writer is procrastinating due to an ugh field around possible rejection.
(I know intelligent people will disagree with me on this, so I thought I’d make my reasoning a bit more formal/explicit to give them something to attack.)
Makes sense. I still think the bar should be a bit lower for criticism, for a couple reasons.
Motivated reasoning means that we’ll look harder for flaws in a critical piece, all else equal. So our estimation of post quality is biased.
Good disagreement is more valuable than good agreement, because it’s more likely to cause valuable updates. But the person writing a post can only give a rough estimate of its quality before posting it. (Dunning-Kruger effect, unknown unknowns, etc.) Intuitively their subconscious will make some kind of “expected social reward” calculation that looks like
Because of human tendencies,
social_punishment_for_sloppy_criticismis going to be higher than the correspondingsocial_punishment_for_sloppy_agreementparameter in the corresponding equation for agreement.If
social_punishment_for_sloppy_criticismis decreased, on, the margin, that will increase the expected values of this calculation, which means that more quality criticism will get through and be posted. LW users will infer these penalties by observing voting behavior on the posts they see, so it makes sense to go a bit easy on sloppy critical posts from a counterfactual perspective. Different users will interpret social reward/punishment differently, with some much more risk-averse than others. My guess is that the most common mechanism by which low expected social reward will manifest itself is procrastination on writing the post… I wouldn’t be surprised if there are a number of high-quality critical pieces of LW that haven’t been written yet because their writer is procrastinating due to an ugh field around possible rejection.(I know intelligent people will disagree with me on this, so I thought I’d make my reasoning a bit more formal/explicit to give them something to attack.)
A good solution could be to just not downvote sloppy criticism. No reward, but also no punishment.