Cheating in online exams and Proctorio

[Caracal]

Some students will cheat. These cheaters cheapen the degree for the honest students.  I am NOT sorry that defending the value of the degree for the honest students comes at the cost of disrupting the exams of someone that is acting oddly during an exam.

It's more than disruption, it's about communicating to students that they are under suspicion, because they are supposedly acting "oddly." The problem is that oddly is defined as failing to adhere to a series of or arbitrary rules. Some students might have a perfectly easy time doing this. If you get a little distracted, or have a tendency to look away when you do something, or anything else, well tough luck, you have to be subjected to invasive checks. It's wrong and it can't be justified by appealing to academic integrity.

[clean]

It is not 'looking away'.  It is repeatedly looking in the same place.  It is watching their eyes 'read' that location.
IF students want to look up, that is fine. IF students want to stare at different places, that is fine.  When they keep returning the the same place, and keep moving their eyes as if reading, well, that is something else. 

It doesnt take long to learn what will trigger a proctor to do a scan.

[phi-rabbit]

I have never felt comfortable accusing a student of cheating on an exam based on an AI flagging it.  My sad experience is that even when I do have really hard evidence of cheating, the appeals committee is not necessarily going to find in my favor, so I'm definitely not going to do it based on something as thin as "they were looking suspicious."  I had to use Respondus with webcam monitoring during a year teaching all my classes online and it seemed to cut down on how many essays on an essay exam were clearly plagiarized (versus in a class at the beginning of lockdown in which I gave essay exams without monitoring and they ended up being 50% cut and paste jobs!).  So it did seem to serve as a deterrent to some degree, but didn't eliminate it.  As soon as I was back teaching in person I went right back to having them hand write essay exams in the room, which a lot of them found strange and old-fashioned.

I also teach an asynchronous online class for my university's off-campus programs wing.  This course has historically always required in-person proctoring because the person from my department who first set the course up designated it that way (and I'm glad).  The off-campus programs folks periodically try suggesting we switch it to Respondus "proctoring" so apparently they really would rather we not make students go in for real proctoring if it can be avoided.  I'm sure they think it discourages students from enrolling. 

During lockdown, I had to allow webcam monitoring in the async class as a temporary measure.  As soon as the testing centers where people could have it proctored opened back up, I went back to requiring in-person proctoring.  Other faculty continued allowing the AI "proctoring" when that really should have been retired, and from talking to some of them I can tell that the exam grades they are getting in there are higher than mine.  I know why they probably do it: because increasingly we get students frustrated or annoyed by the requirement to find a proctor (even though the U. is quite liberal in approving proctors) and who plead to be allowed to use Respondus instead.  I tell them there are no exceptions.  Online classes that require proctoring are marked as such when they enroll, so I don't feel any obligation to make it possible for them to avoid doing it.

[clean]

I have never felt comfortable accusing a student of cheating on an exam based on an AI flagging it. 

That is why I use Live Proctoring with Examity for major exams.
I use Respondus Lock Down Browser for low value quizzes.  If the AI kicks it out, I have in the syllabus that the quiz doesnt count.  I drop several quizzes anyway.

[Caracal]

It is not 'looking away'.  It is repeatedly looking in the same place.  It is watching their eyes 'read' that location.
IF students want to look up, that is fine. IF students want to stare at different places, that is fine.  When they keep returning the the same place, and keep moving their eyes as if reading, well, that is something else. 

It doesnt take long to learn what will trigger a proctor to do a scan.

I'd want to know if there's any way to figure out how reliable those "tells" are. I guess somebody could try designing an experiment where you give students some sort of test and instruct some of them to attempt to cheat-could you offer some sort of incentive for students who cheat without getting caught? Has anybody run an experiment like that? Then you could actually have some sense of how the AI and live proctors do at identifying suspicious behavior-and what the rate of false positives is.

Even if it's a perfect system, I still think its a problem. Obviously, Clean is being conscientious and thoughtful about this. The horror stories for this kind of stuff always come from instructors who are callous and foolish. But I think there is a basic problem here with privacy and invasive technology even when its done carefully. You can try to make a classroom into a relatively secure testing venue. Personally, I design my tests in ways so that I don't have to worry about it being airtight. (Let students bring a sheet of notes, use only essay questions, increasingly I'm writing questions that ChatGPT is going to have problems with ) But, even the pretty light levels of surveillance I use for in person tests would translate into the need for a lot of surveillance for a remote test. And the meaning and invasiveness of surveillance in an institutional space is really different than when its being done remotely through your own computer.