4

Today, I got accused of cheating on an exam I took yesterday. It was a lockdown browser exam, my exam as well as several other students exams were also flagged because of the following reasons.

This was a 20 multiple choice question exam, and here is the response I was given,

  • the answers were identical

  • the time the answers were submitted is identical

The vice chair for my university emailed me stating “there is absolutely no way this can happen by chance The odds are 1 in 10^1000 or something to that order”

The vice chair also explained this will go through due process so I will have a chance to explain my side.

I don’t know how to feel about this, I feel as if there is nothing I can do and all my hard work put into this class is going to waste.

What action should I take? How can I respond to this claim?

Note from comments: I did not cheat.

  • 3
    How many were flagged? How many took the exam? Were all the answers correct? If not, what was the overall distribution for everyone on the incorrectly answered questions? Were they answered in the order given on the exam page or otherwise? Were all of the students in the same physical location? How was the timing done? Arrival at the server? Browser host clock? Some internet protocol timer? I have reservations about trusting the clock, actually. – Buffy Apr 10 at 22:57
  • 1
    @Buffy I don't know how many students were flagged the email stated that I, as well as several other students were cheating. Around 30 students took the exam and my score was an 83 therefore and it was out of 110 points. They mentioned that our questions were answered identically, in a similar order, and the time each answer was submitted was identical. They were not answered in the order given on the exam page I believe, as they stated it was a specific order. Every student took the exam at home. The Browser host was called "Lockdown Browser" The exam was done through Blackboard. – VimRayyan Apr 11 at 0:03
  • 1
    @DmitrySavostyanov I was not cheating, but the statistics say otherwise, in fact I strongly believe all my answers were not the same as others, I don't know how they are making that claim. – VimRayyan Apr 11 at 0:07
  • 1
    @CJR I think the issue with this is that statistics are proven against therefore even if we say we didn't cheat on the exam, we can still be held accountable. – VimRayyan Apr 11 at 0:08
  • 4
    Solving tasks in the same order and with the same wrong answers could both be caused by a subset of the students being confused on the same issues in the exam. – Patricia Shanahan Apr 11 at 23:39
15

If you truly did not cheat then there will be a way to prove your innocence because false claims eventually lead to contradictions. The statistics aren't very threatening because it sounds like your accuser is probably assuming that each multiple choice question is a random event to get a ridiculous probability, but that's nonsense. Exams are written to be answered correctly by those that are prepared. Thus in reality we expect many of those questions to be answered correctly and identically. You can offer to justify your answers to show you answered them on your own.

The identical submission time is concerning, but there is no reason to assume Blackboard is a perfectly coded software. Even programs like Microsoft Excel that are written by experts and probably used by millions have silly bugs in them. If I coordinated with my friends to the best of our abilities we could never click the submit button with identical submission times because we have different response times. If anything the identical submission time better supports there being a bug with blackboard than anything else. You should ask the accuser to explain how the identical times is evidence that you cheated and not that there was a bug with blackboard or any other aspect of online examination. You should also ask your accuser for how long have they been doing examination like this because likely it's new to them and they should be reminded of that.

Good luck.

| improve this answer | |
  • 7
    Hi Cell. Welcome to Academia! When I read the question a bug in the software was the first thing that came to my mind too. Good answer. :) – The Doctor Apr 11 at 0:58
  • 2
    Trying to get equal submission times would be really silly, like waving a flag saying "We are collaborating". It is much, much more likely to be a glitch in the software. – Patricia Shanahan Apr 11 at 3:47
  • 4
    In calculating the probability of the same order they probably assumed each student rolled dice to decide which question to answer next, rather than, for example, answering the easiest questions first. – Patricia Shanahan Apr 11 at 5:58
  • 5
    "If you truly did not cheat then there will be a way to prove"..a lot of people are innocently convicted by courts. I don't know why it should be different for unis.. – user111388 Apr 11 at 8:08
  • 4
    I don't think that equal submission times per question is achievable even if you tried unless the clock is extremely coarse. Human latency says no. Network latency says no. I doubt even a bot could achieve it. I think the clock was broken. Maybe the server is broken. Simultaneity across the web is an extremely difficult thing to achieve. – Buffy Apr 11 at 10:08
6

One way to approach this, in addition to firmly maintaining you did not cheat, and being prepared to explain why you picked the answers you did, is to ask for the assumptions underlying the statistical analysis, and see how realistic they are.

If the instructor cannot state precisely the assumptions underlying the analysis, the analysis is meaningless and should not be used as evidence of anything.

If they can state the assumptions, you should examine how realistic they are.

For example, there may be an assumption that questions that are not answered in exam order would be answered in independent random orders. I have never taken dice into an exam room, or seen a student rolling dice, to decide which question to answer next. There are only a few reasonable strategies, and two students who both have a misunderstanding about the material that makes particular questions look easy or difficult are likely to answer in the same order, as well as getting the same questions wrong.

Another possible assumption is that all wrong answers are equally plausible. Constructing plausible but wrong answers is difficult. I rarely read a multiple choice question without immediately eliminating some of the wrong answers. The probability of two students getting the same wrong answer increases as the number of plausible wrong answers decreases. In the extreme, there is only one plausible wrong answer and a misunderstanding that eliminates the right answer will cause the student to pick that particular wrong answer.

| improve this answer | |
  • 1
    ... and furthermore, if the cause of the misunderstanding is in the course, it is likely that those who do answer wrong, do also pick the same of the plausible wrong answers. At least for the better multiple choice questions where plausible wrong answers are used to test for particular misconceptions. – cbeleites unhappy with SX May 2 at 18:58
3

the time the answers were submitted is identical

That is not how cheating works. Assume we have a cheating scheme. Person 1 (P1) solves the question, marks it, then opens up a communication line to P2, communicates the message, P2 reads it, marks the answer. This is significant amount of delay, I reckon at least 5 seconds. It also assumes the cheaters are communicating the answers one by one whereas in most cases the source of answers will try to secure their test first before helping others. For the answer to be timed identically you need to use some obscure software that will automatically mark the answers on P2's browser when P1 marks theirs.

I think what is going on is that the accuser is being slightly paranoid and is in grave need of a course statistics. That ratio is simply absurd. Lets observe some facts and lets assume "indentical" means +- 5 seconds,

  • Good majority of people do questions in order
  • People's reading speed in a bell curve
  • People's response time is a bell curve
  • People will have the same reading time + response time will answer at similar times
  • Chances of one set of answers appearing is not 4^20. It is only if you are guessing. But grades come out in bell curve. In a test (especially in an easy one) a lot of people will have identical papers simply because a lot of people will nail the easy questions and even in harder questions, usually there will be common mistakes (i.e. people will not pick at random)

Long story short, if you are an avarage person you might get flagged easily in a class large enough.

| improve this answer | |
  • ... and there is at least one quite natural order in addion to starting with the first and finishing with the last question: answering in order of difficulty. We were actually taught (told by our teachers) to tackle exams that allow that in several passes: first all questions one could easily and immediately answer, then going on in order of difficulty/amount of work: this was to ensure one got as many points as possible even if running out of time in the end. I'd think it quite common that many students think the same questions easy or difficult... – cbeleites unhappy with SX May 2 at 19:07
1

My first message is that if you get a chance to contest the "ruling" that you argue only for yourself and that you leave others who are being treated similarly to argue for themselves. If you didn't cheat, just be clear and consistent that you did not cheat and that some other phenomena, unknown to you, must be responsible. Full. Stop.

Issues about multiple choice tests

Good multiple choice tests are hard to design and not everyone does it well. People make mistakes in phrasing and the phrasing can confuse test takers.

Presumably the students all had the same instructor. If there was anything in the instruction process that left the students in some ambiguous state, perhaps subtly misunderstanding some point, then many of the students will have precisely the same misunderstandings. In a course delivered over the internet, with lessened student-teacher interaction this can be expected to be higher than normal. If any question on a resulting exam touches on that topic, then many students will be led to answer incorrectly, but similarly.

So, a group of students turning in identical multiple choice exams is not, of itself, evidence of misconduct. First you have to weed out all of the correct answers. Next you have to account for similar misconceptions and misreading questions with poor phrasing, etc. Only the remaining ones have any validity at all, and, even when a test is well designed, students may be subtly led to the same answers. And your suggestion that you answered about 3/4 of them correctly reduces the set on which any conclusions might be drawn.

Students are often taught to go through all the questions on an exam and answer only the easy ones (or none at all) on the first reading. If students actually do this, and the easy questions are distributed through the exam, it would be natural for students to answer questions independently in roughly the same order as their peers, again, since they had the same instruction.

I don't know of any actual research on this phenomenon, but, importantly, I doubt that there is research that refutes it by suggesting that students generally answer questions in precise presentation order or random order. Low hanging fruit first. Then make a second pass.

But to say that a group of students answered the questions in the same order is, itself, evidence, much less proof, of misconduct, is a stretch.

I think it is a serious error to try to take processes, however good (and I doubt that multiple choice tests are in that category), that work in face to face instruction with proctored exams and simply try to move them to a completely different environment for which there is little research and little instructor experience. To layer on that an assumption that some intervening "sophisticated" technical system will simply "work" and also reliably tell you things that you couldn't know in a standard environment, is, in my view, preposterous.

I would, as a CS professional, want to know how the technical system was tested and verified and whether there is any provable validity at all to making the assumptions described by the OP.

Issues about timing and internet servers

There is some question about whether the OP really means "identical timing" or merely "very similar timing". I don't think that even a bot trying to achieve such a result could be successful due to internet latency. The internet is fast, but it isn't instantaneous. A very coarse clock, one with a tick-time of several seconds, might show patterns, but they would be imprecise at best. I'll assume that the students were not all on the same local network and were distributed. Certainly a bot spoofing a set of students from a single location could achieve very close, but not identical, times. To spoof several, it would need to fire several transactions and each would have its own latency. Some of the requests might actually interfere with others.

But if you add in human latency the problem of achieving near simultaneity becomes impossible. Suppose some individual - the best in class - is solving the problems and providing answers to everyone else. They send out an answer, but very most likely, would do so after entering their own. What would induce them to wait? Why would they try (or even think about trying) to achieve simultaneity of answers. The answer is received by the cohort. The incoming message has to first be recognized by the individuals (latency), they have to find the question, possibly by scrolling (latency). They have to enter the answer at human speed (latency) and then submit it. Then, assuming that they are not co-located on the same network, individual net requests go out (latency) and picked up by the server (fast, but not simultaneous - I doubt the server runs on a Cray parallel machine). Servers cache things, which can confuse arrival times which must be entered (sequentially - again, fast but not simultaneous).

Caveat: One can remove most of the human latency issue by using something like Zoom or a Skype conference in which everyone is able to watch one of the cohort in near real time. If all members stay synched on the exam page then when one makes a move, the others can all answer quickly after the first. If the "lockdown browser" is effective then Zoom would seem to require a costly setup for the cohort.

If my assumption about not being co-located and not being a bot in operation are correct, I conclude that the timing values are spurious data and you should look to the operation of the software (the server) to account for it. As Harry Potter might say: Riddikulus.

Conclusions

Something is wrong. The exam structure itself is (IMO) flawed. The imputed "data" about timing is deeply flawed. I doubt that it is reproducible. Drawing conclusions of misconduct from it would be wrong.

This is why I have, in several posts here, suggested that we need to rethink student evaluation, among other things, in this new environment. And the caveat stated above, if Zoom or Skype could be used in that way suggests an even more important reason to update methodology from first principles.

But, the bottom line here, is that if you did no wrong, just insist that it is so. Illegitimi non carborundum.

| improve this answer | |
1

I think Buffy is on the right track.

A well structured mulitiple choice test has typical patterns of error because they alternative answers will reflect common patterns of error. So you and the other people may have all had the same misunderstanding or error of reasoning. People do not pick their answers randomly or with equal probability. You would want to know, for example, how many other people had the same errors as you had on each of the questions and also how many other people submitted at the "same time" (whatever they mean by that). You would also like to know the extent to which people not accused varied their order of answering including how many use similar order to you.

You might also look up the "prosecutor's fallacy" which is about how people can misuse probability to make it look almost impossible for someone to be innocent. https://en.wikipedia.org/wiki/Prosecutor%27s_fallacy

Addition: So to be clear, since they are making a statistical and probabilistic argument, you need to be prepared to make a statistical and probabilistic argument in response (in addition to denying that you did it). In this case you need to insist that they share the basis for the probabilistic argument they are making, the relevant(de-identified data) and then you will be able to make a counter argument based on data.

Addition 2:

You got 5 answers wrong, presumably the other people got the same exact 5 answers wrong. You all got the other 15 correct. One way to think about this is that each question is independent of the others (being successful on one does not predict being successful on another) and that you have .25 chance of getting each one correct (assuming 4 answer choices). Also assume you are guessing randomly. The chances of getting 2 identical samples of 20 would be very low. This would, however, not be a reasonable model. Presumably you are not selecting randomly (otherwise why cheat or study), and also each question does not have an equal chance of being answered correctly (some are easy and some are hard). Further, each wrong answer is not equally likely if people have studied for the test. If the 5 questions you got wrong were also the 5 hardest on the test (if say only 20% of the students got them right) that is a very different question than if they were questions 80% of the students got right. Also I would want to know what proportion of wrong answers agreed with your answer (and if your answer was one that displayed a common error). I would really ask for a very specific set of data on the questions.

The more complicated question is about the order you submitted questions in. You would want to get the complete data of the order students worked in to understand what the patterns were and whether your pattern was actually very different from the pattern of other students.

I believe it is already explained in other answers that you need to also get information on what they mean by submitted at the same that is separate from the ordering question. Obviously if you are doing questions in the same order there is going to be correlation of time submitted.

I'm a hard liner on cheating, but you do deserve due process and not some hand waving probability analysis with no explanation. You should certainly follow whatever appeal process is available to you.

| improve this answer | |
  • Your answer missed that the questioner is the accused, not the cheating inviligator. What should they do concretly? – user111388 Apr 12 at 18:33
  • No, I explained what data they should ask for and what argument they can use (pointing out the prosecutor's fallacy). I wouldn't tell that to the investigator. The "you" is quite clearly the OP. – Elin Apr 12 at 21:38
1

One of the most straightforward ways to prove your innocence is to simply submit the work that you had done on the side while answering the questions. That should be able to provide an accurate idea of what was going on through your head while you were answering the question (given that it requires calculations). In the case that the questions did not have any work to go along with them and they were completely memory-based, you can bring up your previous record of marks to show that your mark on this exam is consistent with your previous marks. Another way, as mentioned by others, is to sue the university if they do deal with any punishment wrongfully. Of course, this last method should absolutely be a last resort.

| improve this answer | |
0

To be honest, the situation seems a bit suspicious if you both didn't score at least 18/20.

Assuming a Gaussian curve on the score distribution, having the exact same 3 wrongs with the same wrong answers out of 20 questions is highly unlikely unless the class has around 200 people who took the exam.

What action should I take? How can I respond to this claim?

You should object the decision by a written complaint, writing that you did not cheat. And unfortunately, that is basically what you can do.

I agree that this is a terrible examining system, vulnerable to cheating, and probably does not actually measure knowledge (with 20 multiple-choice questions). However, this is what you got, unfortunately. The chances are, at the end of the day, you might end up in a court defending your case if nobody believes you because of the reason I stated above.

| improve this answer | |
  • 3
    Why would you assume this to be Gaussian? – Tobias Kildetoft Apr 13 at 6:49
  • and moreover this argument has a whole lot of independence assumptions that IMHO need a lot of substantiation: IMHO, mistakes are unlikely to be independent between students of the same course (and as PatriciaShannahan explained) also different wrong answers on the same question may not be independent of each other. Also, IMHO not all orders of tackling questions are of equal probability, and FIFO is not the only one that I'd expect to stick out. – cbeleites unhappy with SX May 2 at 19:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.