The most obvious way to test whether A or NOT-A is by having us debate.
These parts are wrong. Debate is not for testing whether A or NOT-A is true, it is for testing what the most accurate posterior for Pr(A) is, given the evidence available, and who had better-assigned priors.
The reason debate is the most obvious test of rationality is Aumann’s Agreement Theorem. If we debate beliefs, and we are both perfectly rational, we will agree on all beliefs debated by the end of the debate. The person whose beliefs pre-debate most closely match the beliefs post-debate, if the debate was strictly rational (rather than using Dark Arts), was the more rational on those issues, and can be presumed to still be more rational on other issues.
So… if you and I debate issue X, and at the end of that debate your beliefs are completely unchanged, whereas mine have changed slightly, then we’ve determined that you are more rational than I with respect to X, and therefore probably more rational than I with respect to other issues… provided that the debate itself is “strictly rational.”
Yes?
If so, two questions: If the debate was not strictly rational, does the debate tell us anything about which of us is more rational? Can you point me at an actual example of a strictly rational debate?
As previously mentioned, there are many other things which are better for being convincing but not rational, so an actual rational debate is pretty much an idealized thing. Some of the early Socratic dialogues probably count (I’m thinking specifically of the Euthyphro). I haven’t read the Yudkowsky/Hanson AI FOOM debate, it might as well.
OK, I’m now pretty sure you’re serious.
So, let me make sure I understand your position. If you believe A, and I believe NOT-A, then on your account all of the following is true:
The most obvious way to test whether A or NOT-A is by having us debate.
If you convince me that A, then you have the more correct belief, and are therefore more rational.
Thus, debate is the most obvious way to test rationality.
Have I understood your position correctly?
These parts are wrong. Debate is not for testing whether A or NOT-A is true, it is for testing what the most accurate posterior for Pr(A) is, given the evidence available, and who had better-assigned priors.
The reason debate is the most obvious test of rationality is Aumann’s Agreement Theorem. If we debate beliefs, and we are both perfectly rational, we will agree on all beliefs debated by the end of the debate. The person whose beliefs pre-debate most closely match the beliefs post-debate, if the debate was strictly rational (rather than using Dark Arts), was the more rational on those issues, and can be presumed to still be more rational on other issues.
OK, thanks for clarifying your position.
So… if you and I debate issue X, and at the end of that debate your beliefs are completely unchanged, whereas mine have changed slightly, then we’ve determined that you are more rational than I with respect to X, and therefore probably more rational than I with respect to other issues… provided that the debate itself is “strictly rational.”
Yes?
If so, two questions:
If the debate was not strictly rational, does the debate tell us anything about which of us is more rational?
Can you point me at an actual example of a strictly rational debate?
As previously mentioned, there are many other things which are better for being convincing but not rational, so an actual rational debate is pretty much an idealized thing. Some of the early Socratic dialogues probably count (I’m thinking specifically of the Euthyphro). I haven’t read the Yudkowsky/Hanson AI FOOM debate, it might as well.
Ah, gotcha. Now that I understand what you meant by “debate”, your position is clearer. Thanks.