Giving unwanted advice seems to be a risk of knowing a lot of stuff - you're going to be able to remember by pattern matching something fascinating and likely completely inappropriate, in response to even the most intimate self-disclosure.

Excellent thoughts.

Quoting from here:

One day a mother came to Gandhi with her little boy for help. She asked Gandhi, “Please, Bapu, will you tell my little boy to stop eating sugar. He simply eats too much sugar and will not stop.” Gandhi told the mother to leave and come back with the boy in three days.

The mother returned with her son and said to Gandhi, “We have come back as you asked.” Gandhi turned to the boy and said, “Young boy, stop eating sweets. They are not good for you.”

The mother then asked Gandhi, “Bapu, why didn’t you tell my son that when we first came to see you? Why did you ask us to leave and come back in three days? I don’t understand.”

Gandhi said to the woman, “I asked you to return with the boy in three days, because three days ago, I, too, was eating sweets. I could not ask him to stop eating sweets so long as I had not stopped eating sweets.”

I think it is possible that this story doesn't go far enough because its possible to do X and not get a positive outcome from it other than "the moral authority to command X", or do X once and get a positive outcome as a fluke.

The ideal sort of advice, from an epistemic perspective, seems like it would involve explaining the general principle that you think might be useful, and then talk about evidence about your attempts to falsify the advice - times you got hurt by ignoring it and times you benefitted from following it. Maybe even pointing to academic research backing it up.

I think this practice, generally followed, would be likely to create a good memetic environment for generating and selecting for honestly sound planning heuristics.

For my own part, I've never been in a context where I could apply this meta-advice with a full theoretical justification of the method because most of the people I try to give good advice to aren't really positively disposed towards being "meta". But I have found that the general process works reasonably well both to help people and to garner status.

I think it is especially useful to "sharing your failures" because it builds trust and shows that you're not simply trying to one up the other person in a stale status game where advice shades into command and a position of command-giving gives a person de facto political authority. Admission of failure provides ammo for later lowering your status by bringing the failure up again later in a different context, so offering it reveals that you're willing to cede ground in places where the ammo might be used later. If you gain authority nonetheless, it is of a more limited sort.

I like LW a lot, but unfortunately I think that with our karma system and public accessibility we're not structurally set up to properly protect people's admissions of failure from hostile scrutiny by the wider world nor to meaningfully reward them with recognition for actually helping people, but in other contexts, where it's honesty safe to apply, I think this is a pretty good formula for genuinely supportive advice.

Do what you can to actually understand the situation, including the resources the recipient is willing to put into following advice.

Thanks for the meta-advice. I try to do this already to some degree, but will try harder in the future.

Pjeby (if I understand him correctly) has been very good about the way people can reinterpret advice in light of their mental habits-- for example, hearing "find goals that inspire you" as "beat yourself up for not having achieved more".

This is insightful and may deserve being incorporated with your meta-advice above. Something like:

"Do what you can to actually understand the situation, including the resources the recipient is willing to put into following advice. Understand how they will reinterpret your advice and then act on such interpretations."

This advice about advice can be thought of as a suggestion to think consequentially about the action of giving advice.

Upvoted for "status transaction". ;)

Eep, I'm working a post full of advice (based on the OKC thread in the open thread). Any thoughts about writing advice to a group of people whose specific responses you may not know?

Also, you might want to tack on a link to the post about other-optimizing; if I hadn't already read it, and were interested in this, I'd want to read that.