> That is, the drug dog who has been right in the past is one we listen to in the future If we're in a situation where we get good data about what's right, we

> That is, the drug dog who has been right in the past is one we listen to in the future If we're in a situation where we get good data about what's right, we can just do science and it doesn't really matter what authorities or dogs think. The place where you need non-scientific epistemologies is when you don't have good feedback data. And sometimes even when there is data, bad predictors can mutually self reinforce. Racist cops arrest more black kids, more black kids get convicted, that proves racist cops are good predictors of criminality. It's not true, but the data says it's true. It's equivalent to a scientist doing many studies and only publishing the ones that are positive. The stats only work for independent measures but nothing is independent especially when you are using predictions to make policy.