Why is getting it wrong good for science?