Is it your point that “Aumann agreement” is also overused?
Not really, that was poor word choice on my part. Only literally is it overused, in that one excessive use constitutes overuse.
It’s that such words have a warm feel to them, so they are used even when the anticipation controlling/more literal/more technical meaning is not intended. The overuse causes confusion by muddying the meaning, and increases the risk that I will name the way to understand the world and achieving my goals instead of actually understanding the world and achieving my goals.
This type of thing is common because one such overuse is common, “rational”. The specific overuse of “Aumann’s agreement theorem”, the same type of thing, is not common.
I have several times seen it described as a rule that rationalists update towards each other’s estimates, which is distressing. Clearly, they may share evidence and conclude something is more or less likely than either originally thought. A way to make sure one is learning and updating is to avoid using words for ideal methods, lest they cause one to think one is using them when one isn’t.
Are you referring to the virtue that Eliezer calls “the void”
Yes. It’s only belatedly and reluctantly named there so it can be an example of its own point, to explain relationships among concepts rather than try and explain by using labels for rationality.
Not really, that was poor word choice on my part. Only literally is it overused, in that one excessive use constitutes overuse.
It’s that such words have a warm feel to them, so they are used even when the anticipation controlling/more literal/more technical meaning is not intended. The overuse causes confusion by muddying the meaning, and increases the risk that I will name the way to understand the world and achieving my goals instead of actually understanding the world and achieving my goals.
This type of thing is common because one such overuse is common, “rational”. The specific overuse of “Aumann’s agreement theorem”, the same type of thing, is not common.
I have several times seen it described as a rule that rationalists update towards each other’s estimates, which is distressing. Clearly, they may share evidence and conclude something is more or less likely than either originally thought. A way to make sure one is learning and updating is to avoid using words for ideal methods, lest they cause one to think one is using them when one isn’t.
Yes. It’s only belatedly and reluctantly named there so it can be an example of its own point, to explain relationships among concepts rather than try and explain by using labels for rationality.
Agreement is an indicator subject to Goodhart’s law.