I think it’s very important to see that there are at least two different ideas/norms around honesty being proposed. There’s:
[Living out meta-honesty in real life means] stopping and asking yourself “Would I be willing to publicly defend this as a situation in which unusually honest people should lie, if somebody posed it as a hypothetical?”
Which is a suggestion for your standards of object level honesty, and separately there is:
And so he simply suggests that on top of this, you should be absolutely honest about where you’ll likely be honest and dishonest.
The idea that you should be meta-honest. You can think about them completely separately, and at the beginning I found lumping them together to make it harder for me to get why the meta-honesty part mattered.
I could be 100% meta honest (when the code of meta-honesty is invoked), and still have an object level honesty policy that you/EY might consider way too loose.
Hmm, I thought this for much of writing the above, but I just figured out the right way of thinking about it. I’d phrase it as follows.
Eliezer is looking for a replacement for absolute-honesty, and it’s meta-honesty. It’s putting lots of work into being as honest as you can on the object level, and being perfectly honest on the meta-level.
There’s an important sense whereby a ‘meta-honest’ person is just a normal, very honest person who’s thought in detail about what honesty means when talking about your own honesty.
Meta-honesty is not an additional variable, it’s just another point in honesty-space to aim for, alongside things like “be absolutely honest”. It’s not an extra thing on top of being honest, it’s a variant of extreme honesty. If you generally tell lies and think it’s fine, then you’re engaging with none of the types of honesty.
I think this is the right call. Maybe I will try to flesh out why, I think that would be worthwhile, but then again, I already wrote a lot above, and I’m not sure how helpful it was.
I think it’s very important to see that there are at least two different ideas/norms around honesty being proposed. There’s:
Which is a suggestion for your standards of object level honesty, and separately there is:
The idea that you should be meta-honest. You can think about them completely separately, and at the beginning I found lumping them together to make it harder for me to get why the meta-honesty part mattered.
I could be 100% meta honest (when the code of meta-honesty is invoked), and still have an object level honesty policy that you/EY might consider way too loose.
Hmm, I thought this for much of writing the above, but I just figured out the right way of thinking about it. I’d phrase it as follows.
Meta-honesty is not an additional variable, it’s just another point in honesty-space to aim for, alongside things like “be absolutely honest”. It’s not an extra thing on top of being honest, it’s a variant of extreme honesty. If you generally tell lies and think it’s fine, then you’re engaging with none of the types of honesty.
I think this is the right call. Maybe I will try to flesh out why, I think that would be worthwhile, but then again, I already wrote a lot above, and I’m not sure how helpful it was.
I also anticipate I’ll write my own review/commentary on the OP, so mayhaps I can expand more on my thoughts and you can have more to respond to.
I look forward to that :)