I tend to think the most exciting ideas arise where supposedly separate fields intersect. One fruitful intersection is that between ethics and statistics, where a supposed fact-value distinction breaks down. (A classic paper here is .)
Thanks to this other thread, I have been thinking lately about the classification-vs-prediction doctrine. From my perspective, what most severely undermines classification-based approaches to medical decision-making is that they impose a utility function prematurely, and this utility function cannot be the patient’s because the patient has not yet been consulted! Thus, the prime objection seems to me an ethical one, invoking individualism as against collectivist ethics.
How do biostatisticians who are savvy to classification-vs-prediction defend their involvement in prohibitive drug regulation, when this enterprise fundamentally endorses classification over prediction?
- Berry DA. Bayesian Statistics and the Efficiency and Ethics of Clinical Trials. Statist Sci. 2004;19(1):175-187. doi:10.1214/088342304000000044