Finding Limits

 Restrictions imposed due to the pandemic have resulted in adjustments in how classes are run, and expanded use of technology. In the Spring, during the first lockdown, my classes were made entirely remote, but as we had little advance time to prepare, we didn't have remote proctoring set up for giving exams and quizzes. I was surprised that the averages on the tests I gave were not much higher than usual, but there are a lot of possible explanations for that. By the time we got to final exams, some restrictions had been eased, and we had students come to campus for proctored finals. Averages dropped like rocks.

We could conclude that our students were "cheating" during the unproctored exams. Although they had to take the exam at the scheduled time, and had limited time to complete the exam, there was no way to limit access to textbooks or other resources. We also expect that some student copied the questions. Easy to do using a phone. So is searching the Internet for answers if you can use your phone during the test. Students couldn't do that for the finals. But if that is the reason, then why weren't the test averages higher on the midterm tests?

I have an alternate theory for why the final exam averages were so low. I think the students were under stress, having to return to school, and being around teachers they didn't really know. Some of my students are in their first quarter, so they don't know the campus, either. I expect that contributed to the lower scores. But I don't discount the effect of the loss of all that access to course resources during the midterms.

Now we are again giving remote tests, which we didn't do in Summer or Fall. We tried a kind of hybrid lecture, with some students opting to attend lectures remotely, and some live on campus, but tests were all live. That was hard for some students, because of travel times. It meant that on the days of tests, everyone had to be on campus pretty much all day, so there were only a few days when the students who chose the remote option could actually stay away from campus. And then Ohio went Purple. More restrictions were imposed. All lectures were remote, again, and students weren't allowed to be on campus except for the tests. Travel time became a nightmare for some. So this quarter, starting under restrictions, we found an app that works with our computer-based testing program, so we can have electronically proctored tests given remotely.

The app requires a picture of each student for a Facial Recognition app, to ensure each student is taking her own test. Another app records audio and video during the test. An AI program reviews the recordings, and creates a report, identifying any anomalies in apparent student behavior.

It sounds okay. It should mean that students take their own tests, and won't get away with using their textbooks or smart phones during the tests. If they leave the chair, they get caught. If they read the questions aloud, that is recorded, and they get caught. That is because we can't control what is in the room outside the range of the computer's camera, so we tell students they aren't allowed to read aloud.

I occurred to me that this means some algorithm will be getting pictures of all our students. And video and audio of their behavior during exams. I don't know what use an algorithm can make of such data, but I expect it could mean something at some point. What stops the test proctoring algorithm from selling the recordings to the NSA? Could this create a permanent record that could be used to identify any of our students in the future? I mean, the system uses AI. I have no idea how intelligent Artificial Intelligence can be, either now or in the future. I feel like I'm participating in a large project to share huge piles of data about thousands of people with the algorithms that may one day rule the Earth. I might be contributing to progress toward the singularity.

I am also making it safer for students to continue their studies. For now, I think that is the more important consideration. Maybe someday I'll ask what becomes of the recordings and collection of photos, but not today.

A Human Choice

 I know that today, the major topics of conversation, at least in America, are the results of the Senate elections in Georgia and the official certification of the Electoral College vote in Congress, but I am not going to write about those.

I saw a headline on BBC News that seemed relevant to this collection of my thoughts. I write bits about algorithms, without actually looking them up or anything, because I read some time ago that algorithms would be taking over the world, because they self-correct and get stronger, better, more accurate, and therefor more powerful over time. That is probably true, but my experience with algorithms has not made me fear them. For one thing, I read Asimov's short story collection, I Robot, which is not portrayed accurately in movies. The actual stories are about the advance of robot technology to the point that robots control most decisions, and keep the peace, direct commerce, and generally make the world better for people. There is minimal conflict between humans and robots, except in the early stages, when some robots either misunderstand orders or take actions that have social effects that clearly the robots don't understand. But the early programmers set the course so robots have no actual ambition. They just have a job, which is to make the world better, and ends up meaning making the world better for humans, which certainly doesn't harm robots. Robots themselves have no desires or emotions, so they have no incentive to "take over" or anything like that. I can see how that story would not make for much drama in a movie, but it is probably closer to reality than the movie, I, Robot, or Terminator, or any of the others about a robot apocalypse.

Algorithms keep offering me advice and ads that I just ignore because of a complete lack of connection to me. Sometimes I smile a little at the oddities, like the constant barrage of ads for enrollment in the school where I teach. Sure, there is a connection between the school and my life, but the ads are still not going to convince me to change that relationship from educator to student.

Today, a Minister in the United Kingdom announced that exams there would be graded by humans, and not algorithms. I don't know much about the British education system. I have read Harry Potter, so I understand that references to OWLS and NEWTS are analogous to A-Levels in real testing, and maybe another that I don't know about. Without knowing the nature of the tests, I can't offer an opinion of the differences between human and algorithmic grading. If they are essay questions, then there would be large differences, and I would give the advantage to humans. I can imagine creating an expert system capable of grading written exam papers, but I'm not sure how much faith I would have in it. Expert systems have been shown to be better than doctors in making some diagnoses, so they could also, at some point, be better at grading essays, but I would be skeptical without convincing data to demonstrate such superiority. I doubt the British public would be very accepting of judgments coming from an AI construct about English literature or musicology.

I wonder, though, if I am just being stubborn about all this. Maybe the algorithms are just as good, or better, at grading essays or stories, at least in some objective sense, and I prefer a human decision because I'm human, which makes me feel closer to the process that way. I expect comparisons could be made, if someone wanted to put in the effort to have a whole stack of essays graded both ways to validate the algorithm. It might work, even in the public were resistant, and the algorithm would command a lower salary, making the process more efficient and cost effective. But that would cost essay graders their jobs, and, really, doesn't that seem like to high a cost? Algorithms are tools we use to make our economy more productive. We should never lose sight of that.