Created: September 11, 2021
Modified: September 11, 2021
Modified: September 11, 2021
distinction
This page is from my personal notes, and has not been specifically reviewed for public consumption. It might be incomplete, wrong, outdated, or stupid. Caveat lector.- Otter notes August 2020:
- When somebody says that X is good---here X could be love relationships, money, peace, or whatever---it is never fully true. There is no X that is always good, unequivocally. So when somebody says x is good, they are implicitly asking you not to make a distinction between the cases---which may very well be the majority of cases---where it is good, and the cases where it's not.
- And the business of intellectual conversation, developing a theory, is all about what distinctions to make and what distinctions, not to make. You have to strike a balance. If you make no distinctions between anything, then your theory is trivial. It's useless. We'll just say, in all states of the world which are all the same state of the world. Then take the action. That is the only action that exists because we don't distinguish it into any smaller sets of actions. It's not a useful theory.
- On the other hand, if you try to draw every possible distinction. Then you will never be able to generalize if you think that every case is different from every other case. And computationally, you just can't draw every possible distinction, your decision tree can only go so deep. When you're planning, you can only consider so many cases, so many potential choices of courses of action. And sometimes that can have, you know, adding additional branching factor can raise the cost of planning exponentially. So anytime you're building a model. You have to balance between drawing enough distinctions to be useful, and not drawing so many that your model is computationally intractable or statistically on learnable.