Some people seem to have criteria for truth that produce self-sealing beliefs.
But yes, I think it would be interesting and valuable to be able to switch out algorithms for different ones to see how that affects the estimated likelihood that the various propositions and analyses are likely to be correct. If an algorithm is self-consistent, not based on circular reasoning and not easily manipulable, I expect it to provide useful information.
Also, such alternate algorithms could potentially serve as “bias-goggles” that help people to understand others’ points of view. For example, if someone develops a relatively simple, legible algorithm that retrodicts most political views on a certain part of the political spectrum (by re-ranking all analyses in the evidence database), then the algorithm is probably informative about how people in that area of the spectrum form their beliefs.
Some people seem to have criteria for truth that produce self-sealing beliefs.
But yes, I think it would be interesting and valuable to be able to switch out algorithms for different ones to see how that affects the estimated likelihood that the various propositions and analyses are likely to be correct. If an algorithm is self-consistent, not based on circular reasoning and not easily manipulable, I expect it to provide useful information.
Also, such alternate algorithms could potentially serve as “bias-goggles” that help people to understand others’ points of view. For example, if someone develops a relatively simple, legible algorithm that retrodicts most political views on a certain part of the political spectrum (by re-ranking all analyses in the evidence database), then the algorithm is probably informative about how people in that area of the spectrum form their beliefs.