I dislike grading. Or maybe, I don't really dislike it as much as I'm discomforted by the inherent randomness of the process. I spend far too much time carefully placing the grade boundary so that "deserving" students can pass, get an A, or whatever.

Or maybe, what I really dislike is how grading affects student behavior. Increasingly, for large undergraduate courses, I see an enormous amount of time and energy wasted on the idiosyncracies of the grading process. Despite me repeatedly telling students that it is far more useful to learn than to chase the grade, they do not listen. And why should they? They have been taught from childhood that the letter grade is supremely important; everyone around them thinks the same; and yes, some important decisions (choice of school, job) are affected by the GPA.

In a discipline like CS, an undergraduate student can learn skills that have direct monetary value: programming, data handling, software skills, and algorithmic design. And at least anecdotally, these skills are valued higher than a GPA in an interview process. But how do I set up the right incentives in my course for students to focus on skills rather than the grade?

I distinguish between individual assignment/test evaluation and the grading. The evaluation is essential because it gives students concrete feedback on what they do (and do not) understand. My problem is with the grading process that distils (or rather, reduces) all the evaluations into a letter grade.

Yes, I know, grading has a significant utility. If a student cannot code basic data structures, she cannot pass CSE101. If a student can write complex pieces of code, reason about their efficiency, and think about optimality, she deserves an A. But that leaves the middle, say, 60% of the class for whom the grade says very little. The grades between an A- and a B- represent some foggy area where a random letter is assigned because we do not know better.

I think that we professors hope for ideal behavior by telling students the benefit of having a "learning" perspective, rather than a "grades" perspective. We ignore the incentive system, probably because it is far too entrenched (and too hard to change). Admittedly, I have no good alternatives.

Initially, I made the grading opaque, so that students could not spend time trying to game it. So they ended up spending time stressing about it, and bombarded me with questions. I made the grading transparent by sharing my grading script. So they ended up computing various borderline scenarios, figuring out the marginal benefit of each point, and bombarded me with questions.

Sigh. I do have a bunch of bad alternatives to my current grading system. Since I assume I live in the real-world (which I do), I have to finally give each student some grade.

• Randomize the grading threshold: For the borders between A and A-, A- and B+, etc. I could just choose the threshold at random, maybe using the total score distribution. At least that marginal utility of squeezing out another point would be low. But any form of uncertainty stresses students out. Besides, it does not send out a strong enough message that learning is more important than the grade.
• Letting students choose their own grade: The Fs and As (or maybe A+'s) are decided. Every other student can choose their grade, maybe through an online form. We assume that students will simply choose the highest possible grade, though I'm not sure that's true. I'd be more worried that certain groups of students (say, minorities) would less inclined to ask for high grades.
• Only give As and Fs: I decide the Fs, and give everyone else an A. I think that student behavior would be the best in this scenario. Though I have now devalued the A for the really deserving students. I could decide As and Fs and simply give everyone else a B. But then students would want to know how to get A and I'd be back to where I started.