Juan F. Meleiro

On Grading

I've come up with a system for grading students that should lessen the evils of our flawed educational system. You don't have to agree that it's flawed to use it, though, just agree on some principles.

First: feedback is essential in education. Indeed, the whole notion of *grading* should be focused on providing feedback for students, and be an integral part on the pedagogical process, and provide some societal validation as a side-effect.

Second: voluntary agreement is important. People should be as autonomous as possible; we are not to take away more freedom if we can possibly avoid it given our condition as evaluator.

We should also stipulate that students come into a course voluntarily, and so have willingly submited to the pedagogical contract. That's not always necessarily true, but is also not something that can be solved by choosing evaluation methods.

Besides, here's an important point: we still have the obligation to provide certification of students towards society. We can't pass someone that has not met the minimum requirements for that.

So here it goes.

The feedback principle would lead us towards a model where students write [1] a series of assignments and receives feedback on them, some sort of grading being generated as a side-product.

But there are some problems.

First, in many first-year courses there are more students than a professor can reasonably handle, if given so many assignments to review. Even with teaching assistants (TAs), that's another demand on the system, and doesn't scale (lest we keep hiring more and more TAs). But you know what grows linearly with the number of students? The number of students. The answer seems to be this, then: peer-review. A standard double-blind review process would be ideal.

Note: the professor only decides whether to accept reject the assignment. There are no grades involved.

Given enough assignments (getting proportionally smaller, as their amount increases), the granularity is enough so that grades are not all--nothing, and the thing seems fine: no one is over-worked, feedback is guaranteed, the pedagogical process is respected, and everyone is happy.

But there's a problem.

What if a student does not learn well in such a model? We are taking away student's freedom to choose how to engage with a subject. Perhaps they would learn better by themselves, in which case we should allow them not to do any assignments. That's *their* choice not to have the (standard) feedback.

But we still need to provide certification.

So, here's the twist: add a test at the end of the term. A single test. The only thing that's left is to account for both evaluation methods in a single formula, lest we draconically demand students to chose beforehand their prefered evaluation strategy.

Here's what I propose:

F = (1 - k/K)*T + k/K

F: final score
k: accepted assignments
K: total assignments (including ones not submitted)
T: test score

Let's unpack that. The greated the number of assignments that are accepted, the greater the final score, up to the limit where they are *all* submitted and accepted, whence the score would be 100%.

In case some assignments are not submitted not accepted, the first term begins to have relevance. In that case, the remaining portion of the grade (that is, full grade minus accepted assignments) will be supplemented by the final test.

Here's what I think are good features of the formula:

Besides, the whole system (formula + peer review), has these features:

Note: the assignments could very well be accepted rejected in a traditional way, though the single-handed evaluation by the professor. It's flexible enough for that. It only asks for the grades to receive a *boolean* (yes/no; 0/1) grade.

Do you have any comments? Send me an email and I'll answer it here, probably.


[1]: Here, “writing” is a generic term for producing any kind of work that pertains to the skills and knowledge that are hoped to be produced during the course.