I like that this responds to a conflict between two of Eliezer’s posts that are far apart in time. That seems like a strong indicator that it’s actually building on something.
Either “just say the truth”, or “just say whatever you feel you’re expected to say” are both likely better strategies.
I find this believable but not obvious. For example, if the pressure on you is you’ll be executed for saying the truth, saying nothing is probably better that saying the truth. If the pressure on you is remembering being bullied on tumblr, and you’re being asked if you disagree with the common wisdom at a LW meetup, saying nothing is better than saying what you feel expected to say.
I find it pretty plausible that those are rare circumstances where the triggering uncertainty state doesn’t arise, but then there are some bounds on when the advice applies that haven’t been discussed at all.
a little cherry-picking is OK
I think the claim being made here is that in most cases, it isn’t practical to review all existing evidence, and if you attempt to draw out a representative sub-sample of existing evidence, it will necessarily line up with your opinion.
In cases where you can have an extended discussion you can mention contradicting evidence and at least mention that it is not persuasive, and possibly why. But in short conversations there might only be time for one substantial reference. I think that’s distinct from what I would call “cherry-picking.” (it does seem like it would create some weird dynamics where your estimate of the explainer’s bias rises as you depart from uncertainty, but I think that’s extrapolating too far for a review)
I wonder about the impact of including something like this, especially with social examples, in a curated text that is at least partly intended for reading outside the community.
I like that this responds to a conflict between two of Eliezer’s posts that are far apart in time. That seems like a strong indicator that it’s actually building on something.
I find this believable but not obvious. For example, if the pressure on you is you’ll be executed for saying the truth, saying nothing is probably better that saying the truth. If the pressure on you is remembering being bullied on tumblr, and you’re being asked if you disagree with the common wisdom at a LW meetup, saying nothing is better than saying what you feel expected to say.
I find it pretty plausible that those are rare circumstances where the triggering uncertainty state doesn’t arise, but then there are some bounds on when the advice applies that haven’t been discussed at all.
I think the claim being made here is that in most cases, it isn’t practical to review all existing evidence, and if you attempt to draw out a representative sub-sample of existing evidence, it will necessarily line up with your opinion.
In cases where you can have an extended discussion you can mention contradicting evidence and at least mention that it is not persuasive, and possibly why. But in short conversations there might only be time for one substantial reference. I think that’s distinct from what I would call “cherry-picking.” (it does seem like it would create some weird dynamics where your estimate of the explainer’s bias rises as you depart from uncertainty, but I think that’s extrapolating too far for a review)
I think the comment of examples is helpful here.
I wonder about the impact of including something like this, especially with social examples, in a curated text that is at least partly intended for reading outside the community.