“observations are theory-laden: even to describe an observation, we need the language of theory. ‘This detector will display the number 5.228’ … which detector? Only a detector built the right way—and ‘the right way’ inevitably involves theoretical ideas.”