The good things about credencism:
- can tell the difference between different levels of confidence instead of just saying of two medicines – one used for hundreds of years and studied extensively, one newly invented – “i believe they are both safe”
- probability/betting math and statistics are nice sometimes
- not skeptical
- not infallibilist
The first one is the main issue.
CF has all 4 of these good points too. How you can get the first one while rejecting credences is the hard part that other people have no solution to.
How CF uses math/stats/etc is relatively simple: it’s just another type of argument. The conclusion of a math calculation is an idea that can be used in thinking like other ideas. So it’s allowed. It just isn’t believed to be fundamental to how epistemology works.
How does CF differentiate different confidences or solve a similar problem? How can CF fulfill the same need? It uses pass/fail grades for ideas. So your first thought may be “two ideas that both pass are equal, and two that both fail are equal, and there’s no way to differentiate two medicines that are both good enough to take (so pass not fail) but one is better”.
That’s the basic issue for why credencists think they’re right.
CF’s answer is to stop giving an idea a single evaluation.
If you only have one single evaluation for an idea, then you need degrees/probabilities/credences or you run into the problems credencists are concerned about.
The way out is to evaluate an idea multiple times. It gets more than one pass/fail grade. That lets two ideas, e.g. two medicines, be differentiated.
What do you evaluate ideas for? Normal issues are like “Is it true?” or “Is it probable?” which lead to giving one evaluation to an idea.
CF evaluates an idea+goal pair. Does this idea work at this goal? (To be more precise, you can add in the context and evaluate an idea+goal+context triple.)
CF rejects talking about “the goal” (very common phrasing) in favor of looking at multiple goals. (I think the usual split of attention for goal and solutions is like 10/90 or even more skewed, but I suspect it should be more like 50/50.)
Two good ideas can be differentiated because, although they both succeed at many goals, there are some goals which one idea passes at and the other fails at.
Instead of saying one idea scores higher than the other, it’s better to identify specifically what goals one idea passes at which the other fails at.
This viewpoint leads to some new challenges and difficulties. For example, there are infinitely many logically possible goals. And every idea passes and fails at infinitely many of those goals. And you can’t get a good overall score (like number of passes divided by sum of passes and fails) to compare ideas. You have to decide what to do by choosing some goals to act on and then acting on any idea which passes on all of your goals (ideas which do that don’t need to be differentiated; if unsatisfied then pick some more ambitious goals). That leads to the question of how to choose goals (or the related question of how to choose values).