In my IMPRS lecture today, I admitted in public that when you make a hard decision (that is, when you choose a model, or choose to include or not include a source in your catalog) you are by necessity doing something subjective. After all, Bayes returns only relative probabilities; making a decision on that basis departs from the objective process of inference. You can do better by making your decisions with the help of explicit utilities, but even these don't save you from subjectivity.

No comments:

Post a Comment