What techniques can be used to prevent cheating on programming assignments?

In his talk at PyCon Asia Pacific 2012, Chris Boesch talks about his experience implementing "Tournament-based Teaching". I highly recommend watching the video of the talk as it's full of cool and fun insights. Programming tournaments (unlike programming homeworks) reward students for having an answer that is better than everyone else's. Winners get cool prizes and bragging rights. This may not prevent all cases of cheating, but it does encourage differentiation of answers and penalizes those who share their answers.

One idea is to force them to include the .git directory for a problem set.

Just make them optional. Then people who would otherwise cheat can just skip them and go about their normal lives instead. More importantly, design exams in such a way that people who forgo the learning process fail miserably. At MIT, I've seen this done in two ways, both of which are pretty effective. For easy concepts, impose a challenging time constraint. You can quickly tell if someone is practiced by seeing how fast he/she can solve easy problems. For example, someone who copies their programming assignments has no chance in hell in solving 10 TopCoder Div2 250 point problems in a 1 hour session, whereas someone who consistently spends a few hours/week on practice problems should find it reasonably easy. Both students can eventually solve the problems, but speed and fluidity of execution only comes with repetitive practice. For harder concepts, require insight. Sometimes the valuable experience in doing homework is not the end result (i.e., what the program finally prints), but the mistakes taken along the way. If you can carefully distill an exam to require such insights, the people who copy the solution, and thus miss the process, are screwed. For introductory classes, this will be hard since the "insights" are so basic. For higher level classes, it's definitely possible. In particular, probability classes are brutally punishing for those who just read solutions to problem sets. I suspect for programming classes, the same is true for knowing when to apply DP and recursion. In both cases, it's possible that some people who cheats on homework will do well on exams. But these are the people you least care about - clearly they are competent and your exams are trivial to them. Forcing them to do your homework is not only a waste of their time, but also your TA's time to grade. For people who aren't grossly overqualified, choosing to cut corners on homework diminishes their chances of doing well in the class. If you can eliminate the incentive to cheat sans punishment, even the unethically-inclined won't want to do it. EDIT: I want to add that I do not advocate this approach in K12 education. In the system described above, the student takes the ultimate responsibility for his/her grades. Prior to college, the teacher and student share this responsibility. In under-performing school systems, the hands-off approach might cause a lot of grief and regret.

Social pressure I went to a school where most professors gave timed take-home exams.  As far as I know, very few people cheated.  I think this was because people knew what a privilege it was to have this trust, so cheating became socially taboo.  At Caltech I never saw people copy solution sets or work together on exams.  At other schools this behavior would be widely accepted and openly talked about. If cheating is part of your school culture, you could also make homework a very small percentage of the grade (around 15-20%).  A lot of students will do all the homework anyway, and the cheaters wouldn't receive a large boost to their final grade.

When I taught C++, Java, Network Programming, and Operating Systems, I required my students to submit their programming assignments by email. After collecting them all, I used JPlag (http://jplag.ipd.kit.edu/) to compare the assignments, then examined similar assignments pointed out by the tool manually to decide if one student copied from another. A common tactic students use is to change variable names and comments. JPlag is not fooled by this. Nevertheless, I make the final decision. Exams would be open book and open notes. Students would be required to solve short programming problems or answer questions that could not just be looked up in their books. They would have to think. Usually this would be sufficient to prevent cheating, but in one case I made up 4 different versions of the exam with slight variations to the questions. In that case I gave 3 students a zero for answering questions from the wrong version, so I know statistically that one student got away with it. On average, I would catch 30% of a college freshman class cheating at some point. The rate was about 10% for juniors and seniors.

One professor at my school implemented a policy of personal code reviewing before receiving a grade on a project or assignment. Essentially you still must submit your work on time, but secondarily you have to meet with the TA and you are presented with a question either about a specific block of code (ie. "What does this loop do?" "How is that variable used?") or something more broad like explaining an algorithm that you used. It's simple in nature, but it requires the student to demonstrate a level of understanding. This still doesn't make it impossible for students to cheat and copy each other's code; however, in my experience, students who don't understand the topics enough to write the code themselves won't be able to explain the simplest parts either.

One of our professors in my University uses Moss (https://theory.stanford.edu/~aiken/moss/) tool to compute similarity between codes. Interestingly he/she has archives of students from different semesters. When students from the current semester submit their code, he/she would compare them with their peers and students from previous semesters. Once during my TAship for a course I asked students to design UML sequence diagrams using IBM RSA v8 to capture a case study. IBM RSA v8 generates XML files for UML sequence diagrams. So what I did is I wrote a code that extracts different information from these XML files such as Number of Messages used, Number of Unique Messages, Number of Loops, Options, Alts, Pars, Breaks and so on. Using this information I compared different sequence diagrams from different students and found three students had the same numbers (messages, opts, alts, pars, breaks etc).

I made a program to programming with flowcharts. It's files contains name of OS user who created that file. It's enough to filter extremely lazy students. For common languages the best is just ask what this program do in a random fragment. While they know what whole program does, they are comfused about specific fragments.

Stop with this thing were grades are valued more than learning. I'm seriously depressed. Guess why? University. What's the best part? I love my major, and I love to learn. When I was a child I spent most of my time learning and reading books. I've always been curious and wanted to know more. School killed that. I still use my spare time to learn, but due to the fact that I have 16 courses per year there is not much of it. It's all about the grades. They give you the book and just about enough time to go through it to solve problems that will be on the exam. There is no time to really get to some deeper understanding, there is no time for mistakes and exploring. They teach you that you shouldn't make mistakes, that you need to be perfect. And then, you realize that you're paralyzed. You can't be creative as much as you want, there is always that voice reminding you that mistakes aren't allowed. You learn not to take risks, to play safe, because you need to be perfect. How can you learn when learning is all about making mistakes, exploring, being creative and questioning everything and there is that voice inside your head that's keeping you from freeing your mind to that? What are grades? Negative reinforcement. You are expected to be great, nobody cares and nobody will praise you for having good grades or getting into good school. But if you do have bad grades, you feel like a failure, you feel stupid, and that makes it even harder to try next time, that makes that voice in your head even louder, and leads to even worse results. I procrastinate a lot. I spend 5-6 hours procrastinating and then do 2 hours of work. But, funny thing, when I procrastinate, it's still about studying and learning. The school system seems to be built on the premise that all students are lazy and that they don't want to learn anything. The truth is, everyone is curious, everyone needs to do something, nobody wants to sit and stare at the wall. But most end up doing nothing because of that voice, that stupid voice that won't let you move because you can't allow yourself to fail. It's interesting how schools managed to kill the curiosity and need for information and knowledge. You really need to be an expert to do that, and really need to try. The premise is wrong. The truth is that people like to be challenged and to work.Don't work against that, work with that. Use positive reinforcement instead of negative one. When there is no punishment, they will do their homework alone.