Minimize cheating on a take home assignment

October 6, 2015 11:10 AM

I am teaching an intro class on engineering design. I've decided to offer an optional extra credit assignment so that students that have missed homeworks can make up some of their homework grade with a bit of extra effort. As part of their regular assignments, they've built a widget. I'd like them to test their widget under similar conditions 4-5 times and report data to me. What are some good ways I can make outright cheating on the assignment a bit harder?

I am not actually concerned with using the data collected for anything outside the class - I'd just like to look at the variability of people's widget tests for possibly trying something different for class next year. Still I'd like to at least have some sense that people are actually performing the extra credit and not just making up numbers.

The widget in question needs about 1/2 - 1 hour to be tested outside and is responsive to some environmental conditions. In addition to them collecting the performance data of the widget I was thinking of having them take a picture of their widget outside before and after the test and send me that picture. I was thinking that I could then check the metadata on the picture and at least make sure that they were outside with their widget during the time they said they were. If they would go to the trouble of making the right photographs, it would be very little extra effort to actually test the widget.

Does this on the face of it make sense?

Would it be easy as pie to fudge the metadata on their end?

If I have them embed their pictures in a word file all together, will I still be able to access the metadata on the pictures?

Any other ideas for ensuring that students who have submitted for extra credit actually performed the test?

posted by permiechickie to Education (9 answers total)

does the testing generate numbers? could you use benford's law to check that the numbers are not made-up? you would need more than 5 numbers for any kind of certainty.
posted by andrewcooke at 11:14 AM on October 6, 2015

Could you ask them for the test data and include a component where they have to discuss their process, observations, etc in a narrative format? Having to produce a document that is more personalized & requires reflection might do the trick (and would have additional pedagogical benefits as well).
posted by dryad at 11:18 AM on October 6, 2015 [3 favorites]

Does everyone have the ability to take these photos? I know some people who are poor and don't have smartphones...this task would be hard for them. I like the idea of the essay. That at least requires some effort and is harder to fake.
posted by FireFountain at 11:28 AM on October 6, 2015

You are making an assumption that the metadata will be accurate/complete. What are you going to do if someone submits the assignment and that data is missing or obviously wrong. If everyone used a phone camera, then you might get accurate data. What about someone who uses a digital camera which has the wrong date/time set?

Some students will fudge the data. There's really no set of criteria that will eliminate someone cheating on a take home assignment.
posted by 26.2 at 11:28 AM on October 6, 2015 [3 favorites]

If you're saying that taking the entire amount of data you request will take under an hour, then I would agree that photos of the widget at the test site would be sufficient proof. Consider specifying that the photo is part of the data set, and recommending the photo to include the widget, plus a sign that says "widget Lastname pre-test/post-test, at site X, date/time". If you want to see the metadata, and not just trust the stated time of day plus the daylight conditions, you would want the photos sent as separate files, not just as an image pasted into a document.

However, if I read your question wrong and it takes over half an hour per data point and you want 4-5 test results, then no. I could easily imagine a student doing one or two tests and then making up similar results for the subsequent ones, because that adds up to a significant time commitment (and probably gets boring). If you want to be sure that results 3-5 are not fudged, I'd require timestamped photos after each test, and possibly scale the extra credit in such a way that it wasn't as profitable to lie, (i.e. one data point gets you 50% credit, each subsequent gets you 20-10% more.) so that overachievers can still get full credit, underachievers may still attempt to cheat out of whole cloth but would have to fake photos, but your midrange students can do one point and call it a day without feeling like they need to cheat to be "finished".
posted by aimedwander at 11:29 AM on October 6, 2015 [1 favorite]

Is it that they need to do a 30-60 min test about 5 times (so 2.5-5 hrs total)?
Or is 30-60 min total enough time to do all the needed tests?

Do you want each student to be testing under different conditions, or would it be good to have them all testing under the same conditions?

If practical, I'd be inclined to set up Testing Hours, certain times when they can bring their widget to a designated spot (outside your office window?) and you (or a TA) will check them in or whatever. It would mean more uniformity of environmental conditions too. It depends how many you anticipate will do the extra credit -- my thought was, if it's a lot of students, going through their metadata/verification steps will be a pain, and it would be simpler to do just a check-in system. If it's only a few students, this might not be as true.
posted by LobsterMitten at 11:33 AM on October 6, 2015 [2 favorites]

Let them know what your goal of collecting this information is, and then tell them you'll be performing statistical analysis on the datasets and will want to follow-up with anyone whose data differ significantly from their classmates'.
posted by teremala at 11:35 AM on October 6, 2015 [1 favorite]

Response by poster: Just to clarify - each test takes 1/2 to 1 hour to complete, so if they did 4-5 tests it would be a big time commitment. However, the homework in question that some people did not submit was also a long assignment (1-3 hour commitment), so I'd like the "make-up" to be somewhat in kind in terms of time/effort. Many of the students are doing fine in terms of homework, so I don't think a lot of the students will even find it necessary to do this.

More clarification on the widgets - they were built as part of a team assignment and they are already testing them once and writing a report about their test as part of their regular coursework.

I like the idea of having them check-out widgets during a time period (office hours) to avoid having to mess with metadata, and also the gradation of points (50% for first test, 20-10% for subsequent tests). I think I will ask for students to voluntarily give me their widgets after they do their first (required) test with their team, and then students can check them out to do further tests for extra credit. Maybe I will have some of the statistical analysis be optional for extra credit as well.

Any other ideas are still welcome.
posted by permiechickie at 11:44 AM on October 6, 2015

More of a philosophical thing, but in my experience the cheating tends to be minimized in situations where I explicitly have a short but broad conversation with my students about trust, responsibility, and academic honesty. Creating "gotchya" tests just sets it up as a game where they try to outwit the system, not a moral choice they're making. YMMV, but if you're set on having some kind of testable verification it would at least help if you don't point out to the students that that's the purpose of it.
posted by you're a kitty! at 5:20 PM on October 6, 2015 [1 favorite]