The Sad Loss of In-Person Exams in Many UK Universities

Is it part of a larger cultural movement, and is it wise?

In the last year and a half there have been indications that many UK universities are planning to eliminate in-person exams. I will not comment on my own university(ies) here, but I hear that Southampton University and others may be moving in that direction. This is part of a “the whole sector is moving in this direction and we should too” argument. Did we fail to understand the saying that my mother’s generation used to say, “if your friend jumped off a cliff, would you too?”

We seem to be losing wisdom in society, or perhaps there are plenty of wise academics around, but the least wise and most conformist are gaining power. I don’t know, but this insanity is ongoing as a cultural movement for the past 4 years. I think it is probably part of a critical social justice movement in academia, specifically the equity-of-outcome aspect of that trend. Though I may be wrong. It seems to also be fuelled by a stated goal of reducing the costs of paying staff to attend in-person exams, which is open to being questioned.

It is demoralizing for all involved, whether they conform or not. The key moral being undermined here is skeptical empiricism—the former bedrock of the greatness and usefulness of university subjects.

So I asked around to find out how many universities are banning exams soon. I hear that Warwick is not. I heard that Bristol University may have already done so, and I checked and they have not. It all feels dubious, with lots of bad information and half-true arguments. Most academics facing such changes, already beaten up and demoralized by a redundancy process last year, seem to have no fight left in them to question this insanity.

I hear from a colleague in Australia that most Australian universities have already banned exams, even at the formerly great RMIT. There does seem to be some truth to the fact that there seems to be a sector wide trend, not that we have to jump off the cliff with them. My friend has now taken a job in academia outside Australia, which is a whole other story. My colleagues in the US inform me that they are not losing the choice to give exams.

When I tell people in UK academia that some UK universities are losing in-person exams, I have reactions such as “that can’t be true, that’s terrible”, “that is utterly stupid, yes it is utterly stupid,” and similar remarks. Both these individuals, successful professors (associate and full), have decided to say nothing about it, though.

My American colleagues responded with terms such as “that’s an absurdity.” These psychologists, and I, know how learning occurs from the research on learning and memory, and we have also a lot of experience. Putting our objections into words, though, takes a lot of work. I’ll do some of that work, but as someone working full time after a round of redundancies, and swamped with new commitments to do more, I will need others to help as well.

An Argument to Keep Exams

In a nutshell, the argument is that in the new era of students using AI to make take-home coursework a passive activity, in-person exams are needed more than ever to maintain real learning, and to maintain the integrity of the professions.

In practical terms, eliminating exams from universities will lead to the use of take-home coursework. There has been a remarkable rise in AI usage on the preparation of these assessments, leading to some of the problems listed here.

1.      Exams reduce opportunities for cheating, plagiarism, and copying/pasting from AI text creation programs.

In-person exams are the best way to reduce cheating and plagiarism (from both AI and non-AI sources) to an almost negligible level.

2.      Removing exams will open the door to a “1-Day-Per-Term Strategy”.

Let me introduce what I will name the “1-Day-Per-Term Strategy” that will become possible with eliminating exams. The strategy could involve the following:

(a)    Attend no lectures, do none of the readings, and do none of the exercises given online. This could be accompanied by obtaining full time work and living outside of the university city.

(b)    At the time of the coursework is due, spend two hours doing the following:

a.      Go onto an AI text generator (e.g. https://chatgpt.com/) and ask the generator to create an academic essay on the topic with APA citations and a reference list. (5 minutes of work)

b.      After the first essay is produced, ask the program to rewrite the essay in ways that fix any initial issues. (e.g., “rewrite and add more citations and add more skepticism in the style of the articles written by [name of module coordinator]) (10 minutes of work)

c.      Copy and paste the essay into an AI detector that graders may use. If the text is mostly detected as being human generated (as sometimes happens), then submit the article on Moodle and print a pdf of the AI detection scan to keep as proof that you did not use AI. (15 minutes of work)

d.      If AI generation is detected in the scan, then rewrite a bit of each paragraph throughout the essay. Run the new essay through an AI detector and if clean, print the screen as a pdf to keep as proof it was written by a human. (30 minutes of work)

e.      If the AI usage is still detected, do further edits to more sentences, and then submit to the AI checker and print the proof as a pdf (30 minutes of work).

f.        Total amount of time spent per module per term = 2 hours. Total amount of work required per year = 2 hours x 8 modules = 16 hours (two days). Amount of learning acquired per year = 16 hours of AI skill building, and 0 hours learning in the subject matter of the modules or course.

g.       Result: A BSc in a subject with all the privileges that that comes with, and with no more understanding of the subject that someone without the degree. In psychology, this could lead to a rise in malpractice in the field.

3.      Removing exams will permanently remove the ability to objectively measure learning objectives, and real learning. It will prevent the falsification of the hypothesis that coursework is superior to exams.

The best way to test learning outcomes is either through a standardized test, preferably at the national level, or alternatively in an objective exam that test knowledge in the subject area (whether that be focused on knowledge or ability to write in the area, either one or both). With the elimination of the ability to have exams we will never be able to assess the potential harm AI usage in coursework is having on real and deep learning. In this way, we will never be able to falsify the hypothesis that coursework is superior to exams.

4.      Financial efficiency of exams.

Exams evolved as a way of assessing large numbers of students in a fair way as class sizes grew over decades and centuries. They remain one of the only ways to cost-effectively measure genuine learning and merit in an efficient way with relatively few trade-offs and disadvantages.

5.      Demoralization of academics.

The majority of staff are passionate about real learning, and the transmission of real evidence and knowledge from their field is often an important life goal and purpose in their lives. Teaching the next generation is one of their core values and morals. With the bypassing of real learning in many students who use AI for coursework, their jobs become an immiserating existence of spending much of their time detecting AI usage.  There are two demoralizing aspects to this: the lingering worry that they may have detected AI usage in error, and the persistent concern that they have failed to detect AI usage. Hand in hand with this is the demoralization of assigning grades to essays in which it appears most of the class has at least started out with an AI essay that they have then modified.

The demoralization of academics can occur in both those who speak out about AI usage, and those that stay quiet. The former group may face stress and hopelessness as they try to communicate the cheating epidemic with others. The stress on these academics may result in health problems over time as they witness their beloved fields diminish. On the other hand, the learned helplessness in those academics that have learned to not protest the loss of exams could be both depressing and demoralizing as well. The shame and guilt that the quiet academics may feel for failing to maintain their field, for failing to show intellectual courage, or failing to fight for real student learning, may linger with them into retirement as a core regret in their professional life.

6.      Demoralization and cognitive dissonance of students.

Students will know that they and their friends are all using AI at some stage of the coursework writing process. Many or most of these students will likely notice that this allows for lectures to be skipped, for readings to be skipped, and for learning to be greatly diminished. Many will also believe that they should learn the topics better, and should not use AI in such a way that they learn very little, and thus will be demoralized from their own behavior that betrays their values. Students who do not use AI will be demoralized by the fact that many other students are using it for coursework assignments. Good students who write their own work properly, and spend hours learning and reading, might become demoralized because their great work is not being noticed in a sea of well written AI-generated papers (written by students not working hard). Even students choosing not to learn and use AI instead to complete coursework will be deeply demoralized and demotivated by the process of doing so—it may linger in their conscience for years or decades.

7.      The failure to detect a truly outstanding and deserving student.

As mentioned above, but in need of highlighting here, truly outstanding students may not be detected, or their apparent brilliance may not be fully trusted, in the context of the possibility that they used AI. Even outstanding students who do not use AI may not have their truly outstanding work stand out in a sea of papers written with AI tools. In terms of science, this may make detecting those capable of doctorates and scientific careers more difficult, and also opens the risk of offering doctoral opportunities to students who are not capable of a scientific career.

Other Potential Problems in List Form

8.      The loss of deep learning in students

9.      The reduction in long term learning retention in graduates

10.  The loss of institutional reputation. Possible loss of accreditations years later

11.  The lack of convincing empirical evidence that coursework, post AI, is superior to exams

12.  The probable hypothesis that exams will promote better understanding, learning, long term retention, and transfer.

13.  Many students and parents may choose universities with exams. Reduced enrolment.

14.  Once exams are gone, it will be difficult to bring them back. Loss of culture and learning in future generations on how to create in-person assessments

15.  Universities are partially state-funded and have a duty to society and taxpayers to produce knowledgeable graduates who can go on to pass on subject knowledge to the rest of society. This is especially true in the sciences, but also in the humanities where learning to do original writing is of great value to society

16.  The degradation of the bachelor’s degrees standing with employers.

17.  The quantitative science of memory, learning, and skill transfer would predict that effortful learning is far superior to passive learning. This predicts exams will induce more learning than coursework in which students

18.  The immense and impossible challenges of AI-use detection. (The impossibility of fair grading)

19.  The loss of academic freedom in university teaching.

Summary

It is incredibly unwise, and even cultic (and probably woke: equal-outcomes inspired), to ban exams in the era of AI text production. I say this as a cognitive scientist who has taught a little on artificial intelligence, and on how the field was born in the 1950s with AI pioneers and cognitive psychologists working together. I also understand large language models—I know what they are, how they work, and I do not fear them. I suppose those who want easy assessments and everyone to pass, are going to label us as AI-phobic at some point, so I will predict that one before it happens. I also encourage students to use AI in the learning stages, but to check for the terrible mistakes and slavish conformity to the internet consensus it has had input into its learning. But for assessments, students should not use AI—not even in figuring out an essay structure. It is better for learning if students struggle to learn how to self-generate essay structures, plans, and of course to learn how to write. This, now, is best done in exams.

My branch of psychology is memory and cognitive research, and in the US memory researchers tend to learn the literature on wider learning as well. I did, with help from colleagues such as Nick Soderstrom from the UCLA Bjork learning lab helping in this regard. My training in social psychology can be helpful as well, specifically in obedience and conformity as these trends flood academia.

In my degrees, and then teaching, I learned, and then taught, how learning is enhanced by the testing effect via retrieval practice. I also learned how active struggle in problem solving helps learning and develops the mind and helps mapping from one problem to another. There is a knock-down argument for in-person exams out there ready for someone to write, deeply citing the literature on memory and learning. But who has time to write it? I don’t in the wake of redundancies and taking on as much work as I can to survive the next round of redundancies. Academic culture is collapsing, and I don’t know if it is driven by equity ideologues, or not. It’s killing the spirit of the wisest and the most intelligent in academia. It’s painful to watch such mistakes while having very little power.

We need an article, fully citing empirical evidence densely, in American Psychologist, or Psychological Bulletin, written by experimental psychologists in the memory and learning area. And it needs to be written by someone at the top of the field. These top journals, when it comes to commentary, like prestige attached to the psychologist and the institution where they are. However, are all such psychologists now conformists? Do they also have intellectual courage of a Scott Lilienfeld, a Steven Pinker, or a Richard McNally? Do they have to teach anymore if they have large grants, and therefore do they care? I am not sure anyone is coming to help in this overwhelmingly demoralizing loss of academic freedom and in this unwise student-as-customers-focused and equity-driven cultural revolution in universities.

(photo credit: jobs.ac.uk)