There’s an intermediate step of believing things because you expect them to be true (rather than merely convincing). It’s fully corrected if you use correlates-to-truth over convincingness for the update.
In other words, if you expect the fifth column more if you see sabotage, and more if you don’t see sabotage, then you can reduce that into just expecting the fifth column more.
There’s an intermediate step of believing things because you expect them to be true (rather than merely convincing). It’s fully corrected if you use correlates-to-truth over convincingness for the update.
In other words, if you expect the fifth column more if you see sabotage, and more if you don’t see sabotage, then you can reduce that into just expecting the fifth column more.