Thanks for the long comment. I really appreciate your candor and perspective—I do think I get the fact that other minds don’t work like mine, but you’re right in sniffing out that a lot of that knowledge is top-down and parts of me are still instinctively typical-minding a lot. I work hard to remind myself, e.g. I have triggers on certain words or feelings that cause me to review memories of specific times when my assumptions about what was going on in someone else’s head were blindingly false.
I think I generally agree with you that there’s a large overlap between rationality and therapy, and I’m intrigued by the hypothesis re: mentally ill rationalists; it seems to be pretty plausible.
Here’s my actual plan if someone fails to act in a way that things seem reasonable. Note that this is the “everything but the kitchen sink option,” including aaaaaallll of the steps one might take, and that for smaller disagreements, this can be done as a speed run or stepwise.
Determine whether to follow up in the moment or later based on the needs of the activity, determine whether to follow up in private, in group, or via delegation based on the apparent needs of the person.
Start by asking. What did they think was going on? What were their thought processes? Assume from the outset that people act in consistent, coherent ways, and that basically everyone is trying to make the world a better place.
Try to pass their ideological Turing test. In other words, try to reflect back to them the priorities they were holding and the goals they were attempting to achieve, and keep falsifying my hypotheses until they give a clear endorsement of my summary.
Ask them to model me, in return (note: one important subthread of how the house will run is a check-in along the lines of “is Duncan clear, consistent, and model-able?”). See if they can predict what my priorities were, and if they have a sense of what I’m reacting to. Do not make this some kind of sick high-pressure quiz dynamic … if they shrug and say “dunno,” I’ll just explain.
Try to lay out, from as birds’-eye as possible a perspective, the conflicting goalsets. Point at the causal chains that brought them into conflict, and highlight my model of where things are broken. Ask them if they have a different model/let them update my picture with a better sense.
Form a new plan for the future; explicitly discuss weighing the goals against one another, and how they ought to stack up. Possibly include other people in the discussion at this point, particularly if the defection seemed to have externalities.
Assume that plan failed. Come up with a plausible explanation for why; try to patch the first or second obvious holes. Form an intention going forward.
Check whether reparations need to be made. Hopefully, there’s a standard formula (as in the pushups example). If not, do a similar process of attempting to converge on a good face-saving/balance-restoring action. If there isn’t a clear satisfactory solution, default to a compromise and schedule a future check-in.
Through all of this, run things by others if either party thinks that’d be beneficial. Also consider things like anxiety/introversion, and have the conversation at a deliberate time rather than forcing it if it’s not urgent.
So yeah, in a sense, this might result in stricter expectations and consequences, but not in a blind, top-down way. In situations where there needs to be an immediate response, I’ll take an action/give an order and expect it to work, but I’ll want to revisit any such quick authoritarian moves after the fact, to explain my thinking and confirm absence of undue harm (and apologize/make amends of my own if necessary).
Overall, though, the idea is to build a high trust environment, and trust goes both ways and is easier to lose than to gain. The thing I want people in the house to actually be justified in believing is “Duncan always has good intentions and is making decisions from some kind of a model. He’ll explain when he can, and if he doesn’t, it’s because he has another model saying why he can’t, and he’ll instead explain both models once the thing is over.”
The idea being that I prove trustworthiness in situations 1-8, and people grant me a little leeway in situation 9. But 1-8 definitely have to come first.
Thanks for the long comment. I really appreciate your candor and perspective—I do think I get the fact that other minds don’t work like mine, but you’re right in sniffing out that a lot of that knowledge is top-down and parts of me are still instinctively typical-minding a lot. I work hard to remind myself, e.g. I have triggers on certain words or feelings that cause me to review memories of specific times when my assumptions about what was going on in someone else’s head were blindingly false.
I think I generally agree with you that there’s a large overlap between rationality and therapy, and I’m intrigued by the hypothesis re: mentally ill rationalists; it seems to be pretty plausible.
Here’s my actual plan if someone fails to act in a way that things seem reasonable. Note that this is the “everything but the kitchen sink option,” including aaaaaallll of the steps one might take, and that for smaller disagreements, this can be done as a speed run or stepwise.
Determine whether to follow up in the moment or later based on the needs of the activity, determine whether to follow up in private, in group, or via delegation based on the apparent needs of the person.
Start by asking. What did they think was going on? What were their thought processes? Assume from the outset that people act in consistent, coherent ways, and that basically everyone is trying to make the world a better place.
Try to pass their ideological Turing test. In other words, try to reflect back to them the priorities they were holding and the goals they were attempting to achieve, and keep falsifying my hypotheses until they give a clear endorsement of my summary.
Ask them to model me, in return (note: one important subthread of how the house will run is a check-in along the lines of “is Duncan clear, consistent, and model-able?”). See if they can predict what my priorities were, and if they have a sense of what I’m reacting to. Do not make this some kind of sick high-pressure quiz dynamic … if they shrug and say “dunno,” I’ll just explain.
Try to lay out, from as birds’-eye as possible a perspective, the conflicting goalsets. Point at the causal chains that brought them into conflict, and highlight my model of where things are broken. Ask them if they have a different model/let them update my picture with a better sense.
Form a new plan for the future; explicitly discuss weighing the goals against one another, and how they ought to stack up. Possibly include other people in the discussion at this point, particularly if the defection seemed to have externalities.
Assume that plan failed. Come up with a plausible explanation for why; try to patch the first or second obvious holes. Form an intention going forward.
Check whether reparations need to be made. Hopefully, there’s a standard formula (as in the pushups example). If not, do a similar process of attempting to converge on a good face-saving/balance-restoring action. If there isn’t a clear satisfactory solution, default to a compromise and schedule a future check-in.
Through all of this, run things by others if either party thinks that’d be beneficial. Also consider things like anxiety/introversion, and have the conversation at a deliberate time rather than forcing it if it’s not urgent.
So yeah, in a sense, this might result in stricter expectations and consequences, but not in a blind, top-down way. In situations where there needs to be an immediate response, I’ll take an action/give an order and expect it to work, but I’ll want to revisit any such quick authoritarian moves after the fact, to explain my thinking and confirm absence of undue harm (and apologize/make amends of my own if necessary).
Overall, though, the idea is to build a high trust environment, and trust goes both ways and is easier to lose than to gain. The thing I want people in the house to actually be justified in believing is “Duncan always has good intentions and is making decisions from some kind of a model. He’ll explain when he can, and if he doesn’t, it’s because he has another model saying why he can’t, and he’ll instead explain both models once the thing is over.”
The idea being that I prove trustworthiness in situations 1-8, and people grant me a little leeway in situation 9. But 1-8 definitely have to come first.