February 26, 2020
Despite studying physics for four years, I never did a single experiment.
It’s not that I didn’t like experiments; I loved them. It’s not that they weren’t required on my degree either; A year of labs was mandatory. I managed to weasel my way out of labs, even though I liked them. Why would I do such a thing?
To maximise my Grade Point Average.
Like the real world, labs are messy. Messiness is fun, but to a GPA hunter like I was, messiness is danger. It’s hard to do an experiment exactly correctly. It’s just as hard to get top marks on a lab report. With ambitions of offers from prestigious grad schools, I couldn’t waste tens of hours per week on assignments I wouldn’t excel on. Like any hunter, I kept my eyes on the prize.
In this essay I’ll explain why many undergraduates live and die by their GPA. I’ll cover what GPA is, and what its used for. Next I’ll do some GPA-bashing, explaining how it causes structurally bad incentives for both students and teachers. Finally, I’ll suggest improvements to GPA which mitigate these bad incentives. Off we go.
GPA, or Grade Point Average, is an attempt to make the academic performance of each student legible to the university machine. It tries to measure how well students understand what they’re supposed to be learning.
Different universities have different systems, but GPA generally relies on compressing all the information about a student’s performance on each course down to a letter grade (say A-F). These grades are then translated into numbers. An A is a 4.0, A- is 3.7, and so on, all the way down to an F being a 0. Finally, these number grades are averaged over every course a student has taken, resulting in a GPA.
Universities use GPA to rank students: the worthy few above a certain cutoff might be put on the “Dean’s List”, while students at the very top of the pile might be awarded scholarships and prizes. The plebs at the bottom might be advised to leave. Most graduate schemes and grad schools have a GPA cut-off: Would-be-employees with GPAs below 3.7 need not apply. Even after the first cut, further sorting is done based on GPA.
One nice thing about GPAs is that they scale. Although written feedback better captures students’ progress, it is infeasible for professors to prepare individual feedback for 150 students. Reports written by different professors are also subjective and hard to compare. GPAs seem more objective, easier to compare, and calculating them takes no time at all.
A high GPA opens doors. It did for me. I wouldn’t have been able to do research at the London Centre for Nanotechnology if not for my good grades, and this experience led to other opportunities down the line.
But a GPA isn’t the be-all-and-end-all. It is an arbitrary, low-dimensional representation of a complex set of experiences. Like representing the biodiversity of a marvellous rainforest with a snapshot of the average tree. Students know GPA isn’t the only important thing about their university experience. But because it is the easiest thing to measure, it is the easiest thing for them to manage. Often, their university experience reduces to a string of attempts to maximise their GPA. Instead of seeing a lovely forest, students fixate on one tree.
Intelligence is multidimensional, GPA is one dimensional. Thanks to Brian Skinner for this slide.
Because a high GPA confers cultural status, as well as real awards like internships and scholarships, students try to maximise it. This has two consequences. The first is that students “game” their way through classes and exams. They focus solely on grades to the detriment of their real learning, which is what the grades are supposed to measure in the first place. The second consequence is of even greater significance. Students decide to take easier courses.
Last year, someone close to me was unsuccessful in a scholarship application for a research position. She had excellent relevant experience and stellar references, but her GPA was too low.
This happens all the time. The kicker is that she applied for the same position two years ago, with much less experience but, thanks to easy first-year courses, a 4.0 GPA. Back then she was offered the “President’s Scholarship”. The internship programme was selecting against relevant research experience.
On its own, a high GPA seems to be a terrible predictor of long-term academic success. Graduate students have to wade through uncertainty, exploring uncharted mental territory without any hand-holding. Due to the mass-produced style of university tuition today, a high GPA gives no sign that a student is capable of this.
Students who are good at gaming assessment are like travellers good at getting from point A to B, provided with a map and clear directions. But a graduate student needs to be able to discover point B from point A, with no directions, writing and rewriting the map as she goes along. For this reason, without qualifying metrics, a high GPA might actually signal someone who is unsuitable for graduate research — yet most departments select these students!
Maybe gaming academic assessments is predictive of success in the corporate world. At work, you have to learn how to operate within an existing system in order to get things done. Navigating different modes of assessment in order to maximise your GPA might be a good way to learn how to do so.
Regardless, the second order consequences of maximising GPA are dire for both graduate employers and academics. These days, nearly every company will tritely affirm that “our people are our greatest asset”. The same is true for university departments, where the strength of a department is directly related to the density of ideas generated by its faculty.
Both employers and research groups want to hire graduates who push themselves, constantly stretching their comfort zones to learn new skills. But maximising GPA encourages them to do the opposite. It rewards students who study subjects they already understand to avoid discomfort. The ones who play the school game, giving themselves the easiest ride.
Easy As have become easier over time. It’s a well-known tragedy: university grades are inflating across the board. Worrying about grade competition, fee-paying students pressure their professors to award higher grades. Worrying about student complaints and high failure rates, professors relent. Universities claim that their students are smarter than ever, but there is less and less to split the difference between the best students and the worst. Worrying about hiring top-talent, graduate employers and grad schools increase their GPA cut-offs, and this vicious cycle renews.
Source: The Economist
These GPA hunting students shouldn’t be blamed for their sins. It isn’t really their fault; they are only responding to bad structural incentives put in place by their university.
I would like to see these incentives change.
GPA was invented to measure students’ learning at scale. But by directly linking it to real rewards, like scholarships, internships, and graduate jobs, it has become a target. Instead of measuring progress towards a goal, maximising GPA has become the goal.
When a measure becomes a target, it ceases to be a good measure.
In business, whenever you target something, you’re supposed to keep an eye on metrics which compete with your target. When I worked in banking, we set a target to grow business by increasing new deals. The sales guys worked their magic, and deal flow erupted. It was only later that management checked the value of these deals. In their rush to meet their targets, salespeople offered such favourable terms to new clients that when taking into account marketing spend, many deals were loss-making for the bank as a whole.
Ideally, you could create an “anti-GPA” metric, which would take into account many relevant things that GPA doesn’t measure. This is close to what graduate employers actually do. While assessing GPA, they also look for “leadership experience”, and participation in extra-curricular activities. If you can throw yourself into several hobbies while maintaining a high GPA, you’re more likely to be a good employee than someone with merely a good GPA.
This works for corporates, which usually don’t require domain-specific expertise from their entry-level employees. But sporting prowess and well-roundedness aren’t important in academia. Instead, graduate schools need students with the potential to produce original independent research. They need students who take risks.
There are hard classes and there are easy classes. In terms of effort and learning, not all As are created equal. The first step to modifying GPA is to make this true numerically.
At both universities I have attended, upper-year students could fill their course-load and boost their grades by taking first or second year courses. It would be easy to penalise students for taking lower-year courses and reward students for taking higher-year courses by ranking all courses on a year-by-year scale.
You could cap the maximum grade possible based on the difference between student’s year and course year. For example, if a fourth year student takes a second year course, maybe she could only get 3.0 out of 4.0. If a first year student takes a third year course, maybe he could attain 4.5 / 4.0. This would encourage students to stretch themselves and reduce the number of Easy As.
A better, more data-intensive, way would be to take into account each student’s background. In the era of big data, this would be fairly straightforward. You could map out the courses each student has already taken. This allows you to determine whether they have the requisite background knowledge for the courses available to them.
For a given course, you can imagine a sliding scale: at one end, a student has more than enough background knowledge for the course (an Easy A). In the middle, the student might be adequately prepared for the course. And at the other end the course might be well beyond the student’s current knowledge.
Sick graphic design courtesy of PowerPoint.
Using this sliding scale, you could create a multiplier which tells you about the riskiness of a given course. Maybe a stretch course is 1.3 times as hard as a normal course, while an easy A is 0.5 times as hard. You could construct a risk-adjusted GPA by multiplying this into the GPA calculation.
Most degrees have a standard sequence of courses. In physics, you have to take classical mechanics before quantum, vector calculus before electromagnetism, and thermodynamics before statistical mechanics. The scale would be set so that if you follow the standard progression your risk multiplier would be 1.0 for all your courses. Only by deviating from the popular path could you raise your risk multiplier.
This new metric would incentivise students to leave their comfort zones in search of a higher risk-adjusted GPA.
Any metric is liable to be gamed. The trick is to choose your metric so that the students who game it successfully also demonstrate the behaviour that you want. People with the highest risk-adjusted GPAs would still be good at gaming the system, but to game it effectively they would continually have to push themselves, learn new skills, expose themselves to productive discomfort.
If grad schools selected against this metric, successful applicants would be more motivated, curious, and eager to learn new skills.
Cultures have enormous inertia, and any change is hard. My guess is most current university students would oppose the changes I am proposing. Today’s students are so used to the idea of maximising GPA that any attempt to destabilise this system will be seen as threatening. This is reasonable. These changes would make university harder. Getting a high GPA directly correlates with getting a high-paying job. It would take time for companies to react to a new GPA measurement, so any changes will penalise students who have to live through them.
University administrators may not want to make these changes either. Lower student satisfaction would cause university rankings to fall. These changes would also cause a short-term decrease in GPAs, as Easy As are graded more accurately. This will lead to more students failing, which would lead to fewer fee-paying students, and hence lower university revenue.
Changing to a risk-adjusted GPA would definitely hurt current students. And it might be bad in the short-term for university administrators too. But it would be better for all future students, who will be encouraged to push themselves instead of slouching along. It would be better for professors, who could try out more advanced material in stretch courses. And it would be much better for graduate employers and grad schools alike, who could screen applicants based on how they adapt to new challenges instead of how good they are at avoiding any challenge at all.
It is difficult to think of a more complex idea than a young person’s potential. No single metric will come close to capturing it. But I firmly believe that we can do much better than GPA.
Thanks to Ned Davies for reading a draft of this essay.