Slashdot Mirror
How Easy Is It To Cheat In CS?
Pinky3 writes "The New York Times has an article on cheating in CS at Stanford. Here is a classic quote from one student: 'I wasn't even thinking of how it [sic] easy it would for me to be caught,' he said. One interesting strategy discussed is for the professor to make the final count for more of the final grade each time cheating is discovered. Share your experiences as a student and/or as an instructor."
Just make the punishment for cheating sufficiently harsh. You cheat.. you get kicked out. Simple.
Many years ago, the University of Virginia had a policy where students caught cheating were kicked out, and then their records were burned publicly. Don't know if they do this anymore.
Liberal? Conservative? Compare perspectives at Left-Right
The problem with cheating is NOT reuse, but dishonesty; the right analogy is not to reuse *your company's* code or report, but *somebody else's* report or code, which leads to lawsuits against your company and firing. In many (most?) programming assignments in school, you are given starting code (like this one except for ....)
Nope; Obfuscation isn't enough. Specialized parsers measuring similarity exist:
http://www.wired.com/science/discoveries/news/1998/02/10464
While you can't catch everyone, disguise is not enough. (And not a measure of skill either).
I forgot probably a small bit of the story:
it took a week for the admins of the university to figure out the trick (tells a lot about their abilities or the lack thereof - i bet they were cheaters too lol), thus they lost a week of time, claiming it to be the reason why I had to leave or face trial.
Those people will eventually be discovered, but I every time that happens it weakens the value of the diploma from that school. That's bad for the graduates who actually worked. The schools need to stop this kind of thing from happening - that diploma certifies that the person earned it. Not that they cheated their way through a program. If it happens enough, they should lose their accreditation.
The professor who teaches operating systems at my university runs people's code through an automated cheating detector. The first operating systems project ends up being around 8 lines of code to complete a stub he provided. In a class of 30 people, there will inevitably be a couple of people who produced similar enough code that they get a zero on the assignment even if they may not even have talked to the person who had similar code. After the first project everyone gets scared and writes the most obscure code they can to avoid being accused of cheating.
On another note, I was a teaching assistant for a lab last semester, and caught a few students cheating. Most often I could hold the two assignments up to the light and see that each line lined up exactly. I had a couple of students who would hand in the same assignment in different fonts, but I nailed them for having the exact same typos.
There were some other people I suspect might have been cheating, but they changed variable names and switched around lines if the order didn't matter to the point where I wasn't confident that the similarities weren't just because they were solving the same problems.
Welcome to the "joy" of statically compiled, un-managed languages. Sure, Makefiles can help but ultimately there is no way to be sure that your binaries are correct short of recompiling everything from scratch in the right order.