![Online vs. In-person student scores on same midterm exam over several years [OC]](https://www.byteseu.com/wp-content/uploads/2025/11/58s3nuu2lo3g1-1536x1148.jpeg "Online vs. In-person student scores on same midterm exam over several years [OC]")
This plot shows student scores (in percent) of students taking the same midterm exam since 2019.
Note: 2020 not included because of Covid disruptions. Scores over 100% are possible because of extra credit questions. 70% represents a passing grade while 90% is an "A" or top marks. The vertical line is the median score. In-person students were not allowed notes or books or internet access while online students were allowed notes but no online aids.
Context: The class is a general education social sciences subject at a public university in the United States. ChatGPT was available starting in Fall 2022, but was used widely starting in 2023.
Posted by Stock_Marsupial3591
29 Comments
Data source: student scores on midterm exam from a course I teach. Tools used: R using “ggridges.”
Well it looks pretty obvious what is going on.
My Son spent the entirety of middle school virtually. Did great with fewer hours per day. When he went back to brick and mortar he was ahead and is now half his classes are in the grade ahead. We loved it.
What’s more noticeable to me is not kids getting more marks in online classes, but the fact that the range of students’ marks in in-person exams over years has increased (meaning more kids are getting worse at it). Am I right? Not sure so please lemme know
interesting how despite everything, there is still a good chunk that is ahead of the online group. they always say AI can do it decently but never perfect.
> In-person students were not allowed notes or books or internet access while online students were allowed notes but no online aids.
Well that sounds pretty unfair, why do it in person if you can use your notes online?
You know, we should just move all schools online.
Everyone would be happy except parents. But since the kids are learning more and the teachers don’t have discipline issues to worry about, and we don’t have millions of cars and busses stinking up the atmosphere and clogging our roads, our environment would be better. It’s a win-win for us all!
Wait why can online students have notes while in-person students aren’t allowed them? Isn’t that unfair towards the in-person students who would’ve otherwise gotten a higher grade too?
People are getting better at cheating it seems.
it should be specific if the exam is take online or in person. Taking courses online are different from taking exams online
Shout out wolf frame alpha
This is a really interesting chart… if there’s no proctoring requirement for the online midterm, then these results are expected, but like another commentator said, it seems like the grades of the in person students are falling, maybe because they’re using AI to study?
By allowing students taking it online to use notes, you introduce two issues for a clean analysis:
1. Online students have notes, which likely improved exam scores
2. Smarter students who pay attention to the rules may be more likely to recognize the advantage of taking the exam online and getting access to notes, causing a selection issue
I spent multiple minutes looking at this data trying to make sense of what could cause such results.
It saddens greatly that the very ovious answer…..cheating…..didn’t even hit me until i read the additional context.
https://youtu.be/bl8Z7Dl7P9A?si=S-QY8rrg67AHlCMO
Nowhere near enough data to make any assumptions. The in-person scores were changing a lot too. Without pre-AI data to show the effect of students just doing the “normal” kind of cheating, e.g., notes they weren’t allowed to have, the only inferences that can be made are from the uniformity of the blue curve shapes – a narrower median means… something.
I doubt the formats were comparable enough to make particularly meaningful comparisons in the first place, so a lot of this will be false specificity after a ton of uncorrected co-variables.
I don’t quite understand, how are there people taking the exam in the same year online and in person? Is there different cohorts still? Is anything else the same, like same teachers, same questions, etc? I’m not from America, so maybe it’s a stupid question.
Then I am a bit surprised how the curve for the online participants is more bell curve shaped than the in person group. Ideally, test results follow a normal distribution. Now if we assume more/some people cheat in online exams, I would expect that this results in a distribution deviating from a normal distribution. I could even imagine two maxima because we add up two normal distributions with different means, one for “honest” students that should he similar to the attendance participants and a shifted one for “cheating” students. This would be my naive guess, but clearly the reality looks different. Does anyone have a good idea, why?
As far as visualization goes I dont understand what is happening, there is a bell curve, no axes, but percentage is on X so what is the bubble then?
I have to ask, cause I don’t see anyone else that did. How come the percentage goes up to 125?
maybe stupid question but are the vertical lines supposed to represent mean or median?
Idk, ñooks like they cheated lmaooooooo
I was a kid and a college student too, once. I know what mlst if humanity is made of. These peoole certainly cheated, at leasta good chunk of them.
interesting data – but i think the graph would be better if it had a title and additional information in the image itself. as is, this image is meaningless without additional context.
If they are in fact identical tests I would expect in person scores to go up too. In Uni, we would always ask the older students for their exams. I don’t remember them being identical but it gave us an understanding of what to expect. Pretty much practise tests. If you used the same midterm for the sake of this case study, these kids should be getting perfect scores. I guess its a bit different for social studies than a class like math and physics where theres 1 right answer for the calculation.
> In-person students were not allowed notes or books or internet access while online students were allowed notes but no online aids.
No offense but i think this data is useless then
If you are online how do they keep you from having some one there with a phone. You read the question out loud and they look up the answer for you.
I’m not proud to admit this, but I had to take a couple of online courses when I was in college. I just Googled the answers when it came to taking tests, still barely passed with a C, as it was a very difficult course.
Can really see the improvement in AI each year
I’m not sure if you’re the professor. If so, have you heard of lockdown browser?
Enforcement on this needs to come from the top down. From a student’s perspective, any class with a curve or even a sense of competition (real or perceived), or even just if professor expectations are based on the median student, will devolve to a prisoner’s dilemma. If you’re the sole online test taker/assignment doer, you will be losing out relative to the original allocation.
Additionally, my philosophy is you can’t really undo AI. It has already been incorporated in workplaces. Students now should learn to use it as a tool rather than a replacement for critical thought. You might also consider asking them to submit their dialogue with the GPT and evaluating them on the types of questions they ask it. There’s a large difference between “Hey, ChatGPT I want to see this data plotted in a particular format but I can’t seem to get the axis right, any tips?” And “Hey, ChatGPT write me a Metropolis-Hastings algorithm for this problem: {screenshot of problem}”
I’ve been on both sides of online education, as a student and a professor. It is not as simple as “be honest.” Cheating was rampant in online education before ChatGPT, trust me. It’s just that most students sucked mighty hard at it (I can’t tell you how many students plagiarized but left the plagiarized parts in different font colors 🤦♂️)
Look, I cheated as needed in HS and in college, you had to be clever and whatnot. Now it’s way way way easier to cheat. They are cheating.