You should note which domains make you surprised frequently, and consider explicitly revisiting your models and assumptions in that domain.

You should notice which domains don't have great models, and check that you are surprised there as frequently as you'd expect - or note that you fail to be surprised, or that you treat everything as equally likely in those domains, and consider if you're doing well.