Thursday, August 10, 2006

Topic for Discussion: Money, Grades, High Scores

Darius pointed out to me yesterday that giving grades to students is a lot like giving incentive pay to workers. The ever-eloquent Joel "On Software" Spolsky has recently (and not-so-recently) written about why incentive pay is bad. In short, it insults the worker by implying that they're only trying to do a good job because they want the money (rather than the more powerful incentive of taking pride in their work and genuinely wanting to do a quality job for its own sake), and it also encourages workers to "game the system" to get more pay while actually doing a less optimal job in the process.

Let's suppose that both Darius and Joel are correct. What are the implications for teaching, if the necessity of giving grades takes the students' focus away from the theoretical goal of actually learning something?

In part, this is why I'd like to make my grading system playful. By doing something mildly ridiculous, I'd like students to forget about the grades and concentrate more on the class itself. (I realize I'm running the risk of having students focus more on grades since the system is novel. I'll let you know how it goes.)

I'll take this a step further and tie the discussion to games. In the early days when video arcades roamed the earth, every game had a scoring system and a high score list. But did we actually play the games to get on the high score list, or did we play because they were fun? Did high scores actually make the games less satisfying, by implying that we were just playing for the score, rather than the sheer joy of shooting countless waves of aliens?

I'm not sure if that's the case or not, but games have definitely moved away from scorekeeping over the last decade. (Of the ten games I'm playing right now, only one of them has any kind of scoring system.) The focus nowadays is usually on "beating the game", and even then the player is often encouraged to go back and do it again with a higher level of mastery -- time attacks, higher difficulty modes, hidden secrets, unlockables, and so on. Essentially, games have evolved from letter grades to pass/fail... with the option to do extra work to earn honors credit!

I feel like there's some universal connection here that's just beyond my grasp, something that suggests a better way to evaluate classes than the standard grading system. Any thoughts?


ocho said...

I think I have a novel thought.
As an experiment, why not have students grade other students; just by their own observations of these students who they share the class with...........

Ian Schreiber said...

I actually was considering that for a time, after seeing an article on "playing the final exam":

However, there are difficulties.

First, as noted in that article, students are more harsh in their grading than the professor would be.

Second, I'm not sure that the grades would be accurate, since being competent is required in order to judge competency (this is why, for example, 80% of us think we're above-average drivers :). Sure, the outstanding students may be fit to give appropriate grades, but the mediocre students may be a bit off in their analysis of their peers, and I wouldn't want to unfairly punish or reward students just because they happen to get a bad grader.

Third, consider the implications if we tied this back to money and the corporate world. If you were working at a software company and your salary increase at the end of the year were based primarily on peer evaluation... would you want to work there?

Gilmoy said...

Well, the primary purpose of university grades is for the university itself to evaluate students, and spread them out within some abstract skill space. Many downstream clients rely on this service. Academia itself uses it for grad school eligibility, scholarship/fellowship worthiness, etc. Industry uses it to cherry-pick the brightest candidates. Students themselves are also their own clients, and should take persistently low scores as a pressure to switch majors/careers.

I disagree with Darius's point. Grades and GPAs are a necessary evil because universities must boil their evals down to something, or the post-university clients would be paralyzed by resume essays. An "A" grade isn't an "incentive", in the sense that it's not optional, and the scale isn't pure bonus. You'd have to extend the corporate "non-incentive" side to include salary penalties, and even outright termination/firing to match what an "F" means, to make the analogy hold up. Otherwise, an incentive of $0 leaves you no worse off than workers in most other companies.

I accept Joel's point: incentive pay probably encourages more system-gaming than true diligence. My whimsy version of it would be something like royalty pay over your code's lifetime and installation base. So if your version 1.0.1 cost you an extra 400% time than the other guy, but it lasts 30 years over 2,000 apps and just keeps working, then eventually your code's royalty-pyramid will dominate. But that presumes a mythical nigh-omnipotent arbitrator service that keeps track of these things. From experience, no human boss has the skill set to do it.

As for arcade games, I think scores-per-action and high score lists just reflect what the technology could handle. When coin-ops dominated, machines simply didn't have the Mflops to do open-ended RPGs with lovely graphics and lots of options; shooters and such was state-of-the-art. (Note that the coin-op industry also had some incentive to /not/ maximize your time-per-coin.) I think it's not a coincidence that when PC and console games reached a complexity level where they could support designer visions of broader scope, coin-ops could no longer compete for gamers' attention, and died out.

@ocho: A limited form of peer evaluations is common in HCI clasess. HCI inherently involves small teams who design an application or gizmo with some interface, then plan a suite of user test studies, then conduct those studies. (So there are universal design-of-interaction themes between HCI and games.) For major milestone assignments, within each team, each student assigns her own "effort %" ratings to all members of her team (including herself).

- If everybody agrees that they all pulled their own weight, they'll all give 100% to everybody.
- If they unanimously agree that one guy slacked off, he gets penalized, etc.

They're not directly assigning grades, but each student's average %, adjusted by the instructor's overall opinion, does earn roughly that share of the team's total score for that milestone. In practice, this policy works so well that young new HCI faculty keep adopting it. (My two teammates accused each other of slacking off, but all 3 of us agreed that I did more than my share of the work :)