For any instrumental activity, done to achieve some other end, it makes sense to check that specific examples are in fact achieving the intended end.
Only if you have reason to believe your naive pattern matching of expectations to observation isn’t already updating your expectations about instrumental activity.
Otherwise, your ″privileging the hypothesis″ that you are in fact wrong.
It’s kind of like smoothing in machine learning. It will have costs and benefits.
Only if you have reason to believe your naive pattern matching of expectations to observation isn’t already updating your expectations about instrumental activity.
Otherwise, your ″privileging the hypothesis″ that you are in fact wrong.
It’s kind of like smoothing in machine learning. It will have costs and benefits.