| > Actually, there is a way to decide if something is scientifically better. Not in a general sense, there isn't. "Better" can only be scientifically defined in terms of a utility function, a goal. If you're trying to conduct, copper is better than rubber, if you're trying to insulate, rubber is better than copper. If you're trying to provide adequate healthcare and social protection to people, then a lower margin of error would be better. > All you have to show is that your construct is competitive within the ecosystem of constructs. Competitive based on what utility function? > But there's no new theory here. How do I predict complications based on what factors? What's the new model? The "spectrum"? A spectrum is a scale with escalating and deescalating values as you travel up and down, where jumps in the spectrum are connected to jumps in prediction. As for abnormal and discrete bins, well, the scientific community already has that. What's new to the table? A reformation of language so that we avoid the word "abnormal"? But where's the improved model? I think an admission that the current model is inadequate goes a long way towards motivating the discovery of better models. > Also note that you propose that there's no way to think about scientific or construct "betterness". Yes there is. You can measure by complexity, prediction, explanation, or generalizability. Okay, so you've named a bunch of utility functions. Now do you really want to apply those to this situation? How do we apply these to the question of whether 1% is a negligible margin of error. Let's optimize for those: 1. Lower complexity: "everyone is a man" seemed to work back in the day.
2. Higher complexity: let's subdivide male and female. There are certainly other genetic traits besides X and Y chromosomes that we could include in our definition of sex. (Hint: It's silly to optimize for higher complexity, but why? I propose that the answer is based on your values.) |