Actual bounded experts who are ignorant on other subjects
Wholly ignorant people who purport to be experts on subjects you’re ignorant of
These are two different things. Gell-Mann amnesia seems to address the latter. You’re referring to the former when you warn against assigning general trust values. Assigning general trust values would actual prevent Gell-Mann amnesia. Correct?
If we there was a practical and efficient way to assign general trust values (and regularly re-computer them too!), and we used them, then yes, that might prevent Gell-Mann amnesia.
I’m not against general trust values – maybe it could be practical and useful. But I don’t think there’s any current way to do this that’s accurate enough to be worth doing.
It seems reasonable to be skeptical about general trust values because it seems strictly better to instead trust individuals on specific topics.
I don’t feel like a general trust value is a useful way to think of anyone, even total strangers. I might have something like a general distribution of trust over some set of topics (for arbitrary people), or maybe a few different distributions for different groups, and definitely distributions for specific people. I guess you could consider a distribution to also be a ‘value’. I was implicitly considering ‘a value’ to be more like a single (real) number.
I admit that I’m not sure how true it is that anyone does already use general trust values in some sense.
Actual bounded experts who are ignorant on other subjects
Wholly ignorant people who purport to be experts on subjects you’re ignorant of
These are two different things. Gell-Mann amnesia seems to address the latter. You’re referring to the former when you warn against assigning general trust values. Assigning general trust values would actual prevent Gell-Mann amnesia. Correct?
If we there was a practical and efficient way to assign general trust values (and regularly re-computer them too!), and we used them, then yes, that might prevent Gell-Mann amnesia.
I’m not against general trust values – maybe it could be practical and useful. But I don’t think there’s any current way to do this that’s accurate enough to be worth doing.
It seems reasonable to be skeptical about general trust values because it seems strictly better to instead trust individuals on specific topics.
I don’t feel like a general trust value is a useful way to think of anyone, even total strangers. I might have something like a general distribution of trust over some set of topics (for arbitrary people), or maybe a few different distributions for different groups, and definitely distributions for specific people. I guess you could consider a distribution to also be a ‘value’. I was implicitly considering ‘a value’ to be more like a single (real) number.
I admit that I’m not sure how true it is that anyone does already use general trust values in some sense.