If you’re one of the three people who’ve read my posts, you’ll know that I believe (a) cheating is going to happen, (b) there’s nothing we can do about it. You should also get the impression that (c) I’m not too worried about it. Don’t we always say “Cheaters only hurt themselves”? If we say it, shouldn’t we stand by it?

But academic dishonesty is an issue: we want the “A” to really mean “good work” and not “knows how to defeat webcam monitoring.” The way forward is through:

We have to scrap the notion that exams in ANY form are meaningful evaluations of a student’s knowledge.

If your department has common finals, try this exercise: take the most recent final you can find. Then type the questions, exactly as they’re presented, into Symbolab (or Cymath, or Wolfram Alpha, or Mathpapa). See what grade a free internet app would get on your final exam. If the free internet app can get a passing grade on the final, what does this say about the student who can do the same?

I don’t know what it says to you. But what it says to the student is that your class has the value of a free internet app. And the question we hope they don’t then ask is, “So why did I pay $4000 for the course?”

So what can be done? The obvious solution is making sure that no internet app can solve a question you’ve asked. That’s easy: just make every question require the student to do a Ph.D.-level dissertation. Of course, most students would end up failing such exams, and an exam that fails everyone is a useless exam. (An analogy that’s helpful: If you want to evaluate the physical fitness of a group of people, you could have them run a 100 mile marathon. Those who cross the finish line are in great shape! But the test tells you nothing about those who fail to cross the finish line)

While you couldn’t write Ph.D.-dissertation questions for a final exam, it does at least suggest a way forward. The thing about a Ph.D. dissertation is that it relies on putting together a lot of little things, over a long period of time. It’s a project, in other words.

In the age of the internet, we must move towards project-based evaluation.

The problem is that doing so is very difficult, especially if you haven’t trained your students to do projects. Telling them, three weeks before the semester ends, that their final evaluation will be a project will cause riots, especially this semester when everything else has changed. Be kind to your students: live with the fact that your final exam is effectively going to be an open book, open note, open internet exam, and accept that some students will get 100s even though they should be getting 30s.

But next semester…now is the time to start thinking about what those project-based questions are going to be. What sort of problem requires putting together everything from the entire semester, and can’t be answered by typing the problem into Wolfram Alpha? (Modeling questions are good: maybe they have to gather some data, construct a mathematical model, cross-check it against reality, and then come to a conclusion about its predictions. In a form suitable for a memo to their boss, whose last math class was “Introduction to ENRON Accounting”)

Wait, you won’t be teaching online next semester because by then the stay-at-home orders will be rescinded?

Sure…but the same issues still apply. Again: If your final exam can be passed by someone using a a free internet app, then what is the value of your course? The “silver lining” of the crisis may be that we scrap the idea of an exam once and forever, and find valid ways to assess student learning.

Last time I talked about the impossibility of monitoring students in the online world: if they’re going to cheat, they’re going to cheat, and there’s not a lot we can do about it. Just the other day, I had a student who got an “11” on the first (in-person exam), based on indecipherable scribbled work, get a “90” on the second (online exam), based on letter hand perfect work.

Can I prove academic dishonesty? No.

Should I worry? One of two things is going to happen. Either the student is going to go to the next class, and their inability to do the work without help is going to hurt them…or they’re never going to use the skills the test was evaluating them on.

Still, there’s the hope that we can still do something resembling exams online with some level of integrity. In the longterm, there’s only one good solution: scrap exams entirely, and shift to project-based learning: You pass an engineering course by building a bridge that doesn’t collapse. Of course, that’s easier said than done, and what it really takes is one-on-one interaction with the students. Fortunately, that is the longterm solution, and we don’t have to implement it by the end of the semester.

The short term solution: Lots and lots of words.

Ideally, every problem should be a word problem. But once upon a time, I got into a debate with a colleague over why we didn’t make every problem a word problem. Their response? While they though it was a good idea, the weaker students needed the problems they could do by following an algorithm. In other words, the non-word problems are there so the weaker students have a chance of passing the exam. The student who won’t flinch when you say “Solve 3x + 5 = 35” will be unable to solve “Three times a number, and five, is thirty-five. Find the number.”

However, in the internet world, this is our (short term) solution to apps like WolframAlpha, Mathpapa, Sybmolab, Cymath, and others. They don’t (yet) know how to handle word problems, even simple ones like the one above. WolframAlpha thought we were asking about newspapers:

Symbolab responded in what looks like Old English (Beowulf, anyone?)

MathPapa’s response:

Of course, a student could figure out that “Three times a number, and five, is thirty-five. Find the number” is the same as the problem 3x + 5 = 35, and then run that problem through their favorite problem solving software.

But maybe that’s not such a bad thing. It’s using a human brain to do what a human brain is good at (understanding questions asked by another human brain), while using technology to do what technology is good at (applying an algorithm that requires no logic, thought, or insight).