49

For certain undergrad classes, I assign a take-home exam, rather than the more familiar in-class exam. When I distribute the exam, I remind the students that this is an individual exam, and that they may not work in groups, and then I add "Believe me, I can tell when you cheat". This, however, is a bluff. Unless it is superobvious, I can't tell if students work together; I only tell them I can to scare them into honoring the rules.

Are there any ways, other than lying to the students, to prevent (or at least minimize) this type of cheating in take-home exams?

Koldito
  • 9,036
  • 5
  • 40
  • 49
  • By individual exam, would using say, Mathematica, count as cheating on a problem? – JNS Feb 02 '15 at 14:05
  • 2
    No. Other than working together, they can use whatever resources they can get their hands on. I allow this because, again, answering the questions properly is effectively impossible unless they understand the relevant primary readings that I cite in the question formulation. – Koldito Feb 02 '15 at 14:20
  • 4
    Any chance you can give us the general subject? – Compass Feb 02 '15 at 14:21
  • A sample question goes along the lines of: "look at this funny phenomenon! Smith 2011 has a nice idea about why things work like this. Explain how Smith's idea relates to the topic of Lecture 5". – Koldito Feb 02 '15 at 14:23
  • Ask for real-world examples and applications. – Compass Feb 02 '15 at 17:16
  • 10
    I can't help but wonder if there are figures regarding how many students make money solving exams for others, and how much. I wouldn't just worry about the testees working together, I would also worry that this might be fertile soil for a sort of black market. Probably much cheaper to hire someone to solve one exam than the white market approach of hiring them to tutor you for weeks. – G. Bach Feb 02 '15 at 18:57
  • 1
    I don't think there is any way to prevent cheating on take-home exams. I tried using them, and it was nothing but problems. For example, I had a student steal his lab partner's paper while she was out of the room, copy her answers, and then put the paper back so she couldn't tell anything had happened. The motivation to cheat is very high, and at most schools in the US the penalty for cheating is negligible (apparently due to legal considerations). –  Feb 02 '15 at 21:05
  • @BenCrowell: got any references for those legal considerations? – SamB Feb 02 '15 at 21:46
  • @Koldito: Can you explain why you decided to give a take-home exam in the first place? What do you hope to achieve with it as opposed to giving "the more familiar" exam in class? – einpoklum Feb 02 '15 at 23:07
  • 28
    Tell us the subject. There's a huge difference between plagiary detection in liberal arts vs history vs math. Is the exam essay, short questions, MCQ, proofs...? Do they submit it electronically or on paper? – smci Feb 02 '15 at 23:14
  • @SamB: We were told by the administration at my school that due to legal issues, it was no longer possible to give a student an F in the course for cheating. This is in California. –  Feb 03 '15 at 02:19
  • 5
    You should not bluff. It will be found out. If you want to see one who is not bluffing, you might want to watch https://www.youtube.com/watch?v=rbzJTTDO9f4 – Thorbjørn Ravn Andersen Feb 03 '15 at 09:23
  • @BenCrowell I'd guess that this means that a teacher can't punish a student with a failing mark by themselves, but needs to report the cheating to the responsible officials who (after lots of bureaucracy) can decide on a punishment. – CodesInChaos Feb 03 '15 at 09:38
  • @einpoklum a take home exam can evaluate how well you do on longer, more difficult problems, and how you deal with literature. – Davidmh Feb 03 '15 at 10:04
  • @BenCrowell Sounds like that's just your administration wanting to avoid drama. At my university in CA (one of the CSU's), the minimum punishment for cheating is an F. – Kathy Feb 03 '15 at 15:18
  • 1
    I remember asking the same question in Sweden. The locals basically answered, "who would even consider cheating?". They apparently have a strong culture of do-on-your-own take-home exams (probably due to having to manage remote students during long, harsh winters), including an honor code. That's probably not easy to synthesize (esp. in cultures that emphasize "whatever it takes"). So, just make the exam so hard that even groups will struggle? – Raphael Feb 03 '15 at 15:32
  • 1
    Another point I wanted to make, in addition to my answer, is that when you tell them "I know when you cheat" - you're really telling me "it's common practice to try to cheat in this kind of exams". I wonder if that doesn't exacerbate the problem more. Also - do people actually cheat? Do you know for a fact this actually happens? Maybe you're worried for no reason? – einpoklum Feb 04 '15 at 21:32
  • @Kathy: At my university in CA (one of the CSU's), the minimum punishment for cheating is an F. F in the course, or F on the assignment? –  Feb 05 '15 at 06:21
  • @BenCrowell F in the course. – Kathy Feb 05 '15 at 18:07
  • Make it a practice to allocate time after the exam to go through the exam, and ask randomly-selected students to explain their answer to the class. Students who have supplied a correct answer but are unable to explain it may have copied it from somewhere without understanding it. It is not necessary to have every student explain an answer, but the risk of being called on will usually increase their fear of being caught. – Mick Jun 19 '18 at 07:36

11 Answers11

37

One way to do it: if the assignment has many small questions, you can make it more difficult taking a random draw from a bigger pool for each student, so they are all slightly different. So, any pair of students would have just a portion of them in common. This can appear unfair, but it should even out if you do it many times.

But, if this were a fight, you would be on the losing side; for any strategy you can come up with, someone else would find the way to hack it. You should instead focus on making people not want to cheat.

  • Make the problems interesting challenges, not mechanical tasks. If it involves some creative thinking it is less likely that two students arrive independently to the same solution (and even less to arrive to the same solutions in each exercise).
  • But make them approachable. If they look impossible, it is more tempting to cheat. In a course I took recently, we had to solve an easier version of a problem, and apply it to a more difficult one. We only had to hand in the difficult version, but handing in only the easy part gave also points.

For what I have seen from a student perspective, the more advanced the course, the less likely cheating is, and the more frowned upon by the other students is.

Dave Clarke
  • 31,848
  • 1
  • 105
  • 144
Davidmh
  • 21,215
  • 5
  • 60
  • 105
  • 9
    Randomizing the questions assigned doesn't prevent cheating. A can still ask B for help, and B can still give help. –  Feb 02 '15 at 20:59
  • 4
    @BenCrowell certanly, it just lowers the probabilities. I can give you my solutions, but working an extra set of problems is more difficult. – Davidmh Feb 03 '15 at 08:23
  • 6
    Giving the students different tests is unfair and doesn't prevent cheating. – David Richerby Feb 03 '15 at 10:31
  • 5
    Why is it unfair to give students different tests, if all of the problems were drawn randomly from the same pool? – Trevor Wilson Feb 04 '15 at 00:12
  • 1
    @DavidRicherby I'm with Trevor here. Some questions are bound to be easier than others, but they will also cover parts of the material one student is more used to than others. When there are enough questions, the exams are statistically fair. – Davidmh Feb 04 '15 at 00:20
  • 2
    @TrevorWilson Being from the same pool is completely irrelevant unless every single question in that pool is of precisely the same difficulty. – Tobias Kildetoft Feb 04 '15 at 10:42
  • 3
    @TobiasKildetoft for more or less uniform difficulty, and enough number of questions, it would be perfectly fair. Difficulty is not uniform across students, and the variability they perceive will be much larger than the random selection effect. – Davidmh Feb 04 '15 at 10:52
  • 1
    @Davidmh Sure, 97.5% of the students will get a test that's within two standard deviations of the mean difficulty. And the other 2.5%... – David Richerby Feb 04 '15 at 10:54
  • 3
    @DavidRicherby The other 2.5% will fall asleep just before learning the material that would have helped answering that question that will be in the exam, so they would get a statistically unfair result anyway. – JiK Feb 04 '15 at 16:27
  • 1
    @DavidRicherby But the standard deviation in difficulty will be negligible when compared to all the other random factors that affect the performance on an exam. – Davidmh Feb 04 '15 at 18:30
  • 2
    @TobiasKildetoft I would change "precisely the same difficulty" to "approximately the same difficulty," where the requisite degree of approximation can be relaxed as the number of questions randomly selected for each exam increases. At some point any "unfairness" created by these small deviations is dwarfed by the unfairness that results from completely different exams being given in different terms, or by different professors in the same term (as well as the random factors that others have mentioned.) – Trevor Wilson Feb 04 '15 at 20:59
  • @TrevorWilson that's wrong; it's fair by law of large numbers. you're expecting a lot of perfection in the academic process besides this point, for this to be relevant. for instance, by your same logic, it's also unfair to select a question that one student might happen to have studied better than another student. i.e., literally any question. –  Feb 04 '15 at 21:06
  • @AAA I think you may have directed your comment to the wrong person. – Trevor Wilson Feb 04 '15 at 21:19
  • Yep. /cc @TobiasKildetoft –  Feb 04 '15 at 21:21
22

The solution I've observed at my Alma Mater: Take-home exams are given only with very few questions, which are non-trivial and open-ended. Now, if several students work together on this exam, but every one of them manages to come up with an answer which doesn't read like a copy of the other students' answers - then, well, that means they have some sort of command of the material, even if they didn't come up with the idea themselves.

Of course, this is mostly relevant for more advanced courses. In more basic courses, there are never any home exams.

einpoklum
  • 39,047
  • 6
  • 75
  • 192
  • 3
    This doesn't, e.g., prevent a student from paying another student to do his exam for him. –  Feb 03 '15 at 02:20
  • 10
    @BenCrowell short of assigning a spy drone to each one of your students 24/7, there is no scheme that could prevent that. – Davidmh Feb 03 '15 at 08:25
  • 3
    @Davidmh Not using take-home exams mostly prevents ringers. – David Richerby Feb 03 '15 at 10:32
  • 1
    @BenCrowell: There have not been any known cases of this happening, at least in my department. Also, the hand-writing would not be the same, and the style would be different than the student's homework assignments (if you've saved copies). Now, sure, the culprit might transcribe his contractor's work, but then you just really want the 24/7 spy drone... – einpoklum Feb 03 '15 at 12:33
  • 1
    @einpoklum: Who handwrites anything outside of exam conditions? (I don't handwrite things in exam conditions either, but that is because I physically can't) – Frames Catherine White Feb 04 '15 at 12:37
  • 1
    Well, whoever can hand-write an exam in class can hand-write it at home. If the teacher trusts you s/he might accept a typed-up solution. – einpoklum Feb 04 '15 at 16:01
15

Don't, maybe? They can't cheat if collaboration isn't against the rules.

Do you care what the students know, or do you care how they learn it?

Why not consider setting it as a coursework assignment, and taking away the restriction entirely? You open the door to them being able to collaborate, work together, come up with creative solutions. Tell them to declare who they worked with in a short section at the top.

Education has this strange focus on separating people: when was the last time you did an assignment, research project or similar truly alone? In the professional world I've never once done a truly solo project: finding solutions as part of a team is a vital skill, and this sounds like a perfect opportunity to encourage it.

Grading is important, of course, and there is a time for differentiating between students... that time is in the formal, controlled examinations. Everything else is a learning exercise, and if they learn it by working with a friend, fantastic.

Jon Story
  • 593
  • 3
  • 10
  • 2
    I agree. You can't prevent people working together so rather than penalizing the honest ones it's much easier to allow working together and structure the homework appropriately. – Tim B Feb 04 '15 at 11:23
  • 2
    "Tell them to declare who they worked with in a short section at the top." Even better, ask them to write down, in detail, exactly what each collaborator contributed. – WetlabStudent Feb 13 '15 at 01:57
  • Yes, and in the real world it is often very difficult to tell who has actually done the work and people tell untruths about it to gain advancement. The problem with this idealistic standpoint is that although some students will collaborate in an equitable way, there are others who are happy to copy work and not have the faintest understanding of what is going on. Yes, they will get found out in an exam, and that's fine so long as the coursework component is a minor contributor to the overall mark. – ProfRob Nov 25 '20 at 11:46
13

Spend 10 minutes interviewing each student on part of their answer. You don’t tell the students in advance what part of the answer you will ask them about, and you ask students different interview questions depending on their answer.

If the student can show they understand and can expand the answer, then do you care if they cheated?

The interview is also lickly to help the student learn and provide you with good information on areas that lots of students are finding hard.

Ian
  • 1,160
  • 7
  • 11
  • 4
    ...and if there are 400 students in the class? Even only 100 students and it makes for two full workdays conducting interviews. – J... Feb 03 '15 at 17:15
  • 5
    @J... We use that strategy for software assignments in a class of 600. It's expensive, but it works. It also offers the opportunity for personalised feedback. – sapi Feb 04 '15 at 07:39
  • 3
    @J...: I announce in class that I would interview a sample of the students. In any case, a take-home exam in a 400-student class is not a very good idea. – Martin Argerami Dec 13 '16 at 02:38
9

Some of the MOOC (massively open online courses) such as MITx 6.002x appear to give each participant slightly different parameters for exercises. That prevents students from directly copying results, though they can, of course, collaborate on the solution methods (and that is encouraged via online forums). It would probably be too much effort to do this manually, but perhaps you could have a few versions and distribute the questions randomly (or not-so-randomly if you have specific suspicions).

Spehro Pefhany
  • 204
  • 1
  • 4
  • This is what my university does for physics homework. The traditional problems have different numbers for everybody, and the multiple choice problems have slightly different questions and answers. – Greg d'Eon Feb 02 '15 at 18:33
  • 3
    @Kynit: That only works for the most casual type of cheating, on an assignment where the professor doesn't intend to personally look at the student's work. Changing the numbers won't prevent student A from copying student B's method and plugging in the different numbers. –  Feb 02 '15 at 21:02
6

I recently received a take-home exam with an interesting surprise: 50% of the credit was for a problem that required us to augment our solution to a previous homework assignment. There was no way students could work together on this problem because we had all previously turned in different solutions to that homework assignment, and our instructor graded the exam in the context of our previous homework submission. This may or may not work depending on the format of your assignments (the problems would obviously need to be open-ended).

That said, most instructors at my school only give take-home exams when they do expect students to work together.

Michael
  • 61
  • 1
  • 1
    This might be problematic if some people didn't succeed in their previous assignment – now they are punished a second time. (Assuming failing one assignment doesn't automatically fail the whole course.) – Paŭlo Ebermann Feb 03 '15 at 18:34
  • 1
    One of my professors dealt with a cheating problem during a midterm exam by putting a question on the final exam which required us to refer to our final projects ("write SQL queries of sufficient complexity to your final project database..."). That question accounted for 25% of the exam score. (Note that these were in-class exams, not take-homes.) – bwDraco Feb 04 '15 at 00:22
  • 1
    @PaŭloEbermann Not necessarily; it depends how the "augmentation" is graded. Also, students should know that they are expected to learn to do the homework problems correctly even if they did them incorrectly the first time (although it wouldn't hurt to remind them of this.) – Trevor Wilson Feb 04 '15 at 20:38
  • @PaŭloEbermann ... My comment assumes that the previous homework assignment has already been graded and handed back; otherwise I can see how your objection would apply. – Trevor Wilson Feb 04 '15 at 20:40
5

There is no fool-proof system so the task becomes how to make something that reduces the number of cheating opportunities. In my experience, I see take home's as a tool for advanced courses with smaller number of students and the rest of my response needs to be viewed from this perspective.

A first consideration regards the type of questions asked. Simple questions with "obvious" standard answers are not generally suitable because the answers can easily be copied. Hence essay type answers where no obvious unique answer is possible is better. This hints at answers where the students understanding and knowledge has to be synthesised is the aim. In my field, I have used several images of landscapes asking students to chose one and identify and describe processes as one such type of question. It is quite difficult to cheat n this type since I would not expect two students to identify the same subset of possible observations to discuss. How this type of question can be transferred to other disciplines is a matter of imagination. another favourite is the following

Asking relevant questions is a key aspect of academia. Formulate a key scientific question within "the topic" and provide an good answer to the question.

With simpler type of questions, I believe providing a narrow time frame for replying can be part of a solution. Here there are two main ingredients, one is to provide questions in a random order to the different students so that question n is likely different for each student and then to provide a very narrow time limit for students to respond. Questions can, for example be made available through a server (of some sort) at a given time and then requesting answers to be submitted before a given deadline, either enforced strictly or by a, possibly incremental, reduction of credit by degree of lateness. An alternative is to release questions incrementally to each student and enforce a strict deadline before another question will be issued. I do not have suggestions for how to implement such an exam but it can be accomplished with simple learning platforms but may involve some work by the teacher to facilitate. I would not think it is useful for very large groups.

Peter Jansson
  • 73,217
  • 12
  • 206
  • 344
  • 8
    Tight deadlines should be broad enough to fit into students' lives; otherwise, people may feel more tempted to cheat. Also, you may run into problems with home network suddenly broken. – Davidmh Feb 02 '15 at 16:04
  • +1 for the idea of having students actually write something about their thinking – SamB Feb 02 '15 at 21:44
3

Lock an RFID ankle bracelet to each student, and hand them a tablet and headphones. The tablet contains the test, and the headphones must be worn whenever they are answering test questions. The tablet's camera can verify the face and proper headphone placement. Further, it will determine if there are any other students from this class within visual range via the RFID ankle bracelets. The bracelets detect tampering or covering (such as in tinfoil). The headphones have microphones and detect if there are any nearby voices loud enough to penetrate the headphones, thus defeating using telephone or other means of vocal communication. The face recognition on the tablet includes eye tracking, if the student looks at anything other than the tablet during the test, the question in view is graded 0.

Alternately, don't use take home tests where cooperation can alter the results.

It seems that the only reason to give a take home test when cooperation would be a problem is if the time to complete the test is longer than a single class period. However, there are few times when this should be the case. Break such tests up into smaller quizzes, use a testing center, or any number of other solutions that will allow you to test without sending it home with them.

So the correct answer is - you don't. If the test depends on non-cooperation, then you must have it supervised. If you can't supervise it, you alter it so cooperation doesn't pose a problem.

Adam Davis
  • 3,257
  • 16
  • 15
0

Here's my position on handling cheating on take-home exams.

  • Make it clear that answers must reflect independent effort. I would put this in the syllabus:

    You may consult your textbook, notes, and standard Internet resources when taking take-home exams, but your work must be original and you may not solicit or obtain assistance from or provide assistance to other people for any specific content on the exam. Activities considered cheating include (but are not limited to) copying or closely paraphrasing content from websites, discussing exam questions with other students, and asking for help with specific questions on Internet forums. All exams are checked for originality and copied content and anyone found cheating will be assigned a failing score for the exam.

  • Make sure questions assess actual understanding of the content. Simple multiple-choice questions are easy to cheat on in a take-home exam. Free-response conceptual questions provide a more effective assessment of student understanding. They can be made more resistant to cheating by using questions that cannot be easily answered in their entirety through simple Internet research. In addition, it is easier to determine if a free-response answer is original or copied compared to multiple-choice responses (more on that in the next point).

  • Check exam results closely for cheating. In addition to checking for highly similar or copied answers among exams, I would check responses online to determine if they are copied from some online resource. You can use Google to do this—search for select phrases in answers in double quotes to find exact matches for the phrases. This can go a long way towards detecting cheating in take-home exams.

bwDraco
  • 151
  • 7
0

Take-home exams have always baffled me. I've never really understood their purpose when the material assessed could be achieved through a traditional or open-book exam, or as a time-sensitive (no extensions) class assignment.

Regardless, I did complete a few during my undergraduate career, and looking up old submissions (that yes, for some crazy reason I still have on my hard drive) they tended to be essay style take-homes. This meant that the possibility for cheating was on the lesser side, since it would be easier to detect if a student copied another student due to the subjectiveness of the task.

As others have suggested, exams that have very standard responses are perhaps not the best route to go for a take-home exam. You would be better to have a much more open-ended style that suits the subject you are teaching.

You could have the take-home exam be perhaps a broad question that students have to answer through the development of a project using the knowledge they've gained in the course. After all, it's not the core material itself that is of the uptmost value, it's the level of critical thinking students develop when using said material to attempt to solve complex problems or issues.

awsoci
  • 1,608
  • 12
  • 17
-1

How do you mark the papers? Do you mark a paper in its entirety before moving on? If so this is wrong. You should mark everyone's questions 1's first then all the question 2's etc... This will help you too see any patterns and also keep consistency. What course do you teach? If it is maths/physics then it should be very obvious. I suppose it should be very obvious for essay style questions as well.

Al jabra
  • 295
  • 1
  • 2
  • 7
  • 1
    This doesn't seem like it would be an effective way to detect or deter unauthorized collaboration. It would only catch the most obvious cheating, that the student hasn't attempted to hide at all. – ff524 Jan 10 '16 at 22:22
  • @ff524, in high school my math teacher told us that to be able to successfully cheat you need to know enough to pass without cheating. In my experience, this shows: cheaters are incompetent. – vonbrand Jan 11 '16 at 00:36
  • @vonbrand Are you suggesting that all/most cases of cheating are obvious, and trivial to detect and prove? That's the aspect of this post I was criticizing, and I really don't think that's true. – ff524 Jan 11 '16 at 00:39
  • @ff524, sadly the only cases of cheating I'm reasonably sure of have been bungled cases, in one way or the other. So the sample is certainly biased. Yes, I've learnt of sophisticated cheating arrangements, but they have been far in between. Also note "I'm sure they cheated", and even "I'm quite certain they did it so and so" is certainly no "proof". – vonbrand Jan 11 '16 at 00:52