Today I came across this article (courtesy of Slashdot). I know we are getting better and better at letting computers do work for us, and one day we might have true artificial intelligence. But even though I am not actively keeping up to date on the latest research in this field, I highly doubt that we are at a place where we can leave the grading of papers to the computer.
I was an assistant teacher at my university roughly a decade ago, and I know how hard it is to grade exams. And I only graded CS exams where the problems were short programs written in a very well defined and unambigous language. Even for people this is a hard problem. Just grading can be simplified, but being fair and consistent in grading, very hard. One might object and say that a human grader is prone to subjective grading and a computer would be more objective.
Still, a computer is not better than the programmers programming it. It can learn, but only within the constraints set up by the original programming. If it cannot perceive things outside of a box, due to programming, how can it use that and add that to it’s knowledge?
I remember when I was a student in the course which I one year later taught in. In one homework assignment I had a moment of genius and solved the homework in a non-standard way. The grader didn’t like my solution, even though I clearly got the correct answer and without it being any less efficient or hard to understand. Luckily for me, my teacher which the grader answered to didn’t agree and gave me a much better grade on that homework.
I think it is great that some grading work can be done, but relying on a computer to grade the homework has a big drawback. I think it is neat that we can offload some work and the student can get a quicker feedback, but learning is not only something that the student does from the teacher. A good teacher exists in a feedback loop with the students. While grading the students essays, the teacher learns what the students’ didn’t grasp and where to put the effort. Perhaps a grading system can tell the teacher what areas students need to focus on, but I doubt that it works as well as if the teacher graded the exams himself/herself. Second hand information is rarely as good as first hand.
If you, as a teacher, teaches to many students that you cannot keep up and keep track of their progress, you should seriously consider holding fewer classes and lowering your work load. Being a good teacher means that you are in contact with your students, not that you try to disassociate yourself from them.
The day computers can grade exams reliably for us, it can also write exams for us. Then humanity is in for a bleak future, we are then becoming more and more obsolete.