Blinding us with science.

Restaurant Critic and Bay Area booster Michael Bauer takes a moment to explain how he awards stars:

In assigning stars, I try to start with a fairly objective procedure. Food is the most important component -- in fact, it counts twice as much as everything else. In figuring the overall score, I double the rating for food (3.5+3.5=7 stars), add the ambience (2.5) and the service (3), which gives a total of 12.5 stars. I then divide by 4, which means it comes in a little over 3 stars. While it's the reviewer's decision to assign the final number of stars, when the food is higher than the other elements, we tend to bump up the stars, which is what I did with SPQR.

It's nice to have a system, but you could have the quadratic formula on the back of your tip table, whip out an abacus, slide rule and Deep Blue -- the data Bauer is crunching are still subjective. A couple of the comments allude to this issue, but it is remarkable how frequently this kind of sortilege works. I know, because I do it myself in my day job. Doing Fessering of the humanistic variety, the evaluations I make are subjective. I like to think they are fair, and based on reasonable criteria, but they are not fueled by quantitative data. However, students, especially from the sciences, tend to respond to A,B,C grades as if they were arbitrary and capricious. A 77, rather than a C+, however, seems to carry the weight of truth -- the atomic weight of helium is 8, you got a 77 on your paper. So, the process goes like this. I read the paper, decide on a letter grade, convert that to a decimal grade for the assignment, figure how many points that's worth based on the percentage for that assignment. I add up these numbers at the end of the semester, and use them to assign a letter grade, which is then converted back to a decimal GPA.
    The broader question is if the relation between quantitative evaluations and objective data is privileged. Would it be better if we used language for subjective evaluations, and numbers for objective ones? Dudley Moore might disagree.