Calculated Ease: How STEM Grade Inflation Distorts Harvard's Mission
By Jason Morganbesser
Over the past ten years, STEM at Harvard has become increasingly dominated by a small number of majors–Computer Science, Neuroscience, and Applied Mathematics. These fields went from comprising 37% of STEM concentrators in 2016 to over 48% in 2025. While greater economic opportunities in software engineering likely motivated the shift toward Computer Science, the rise of the other two majors is more difficult to explain. Neuroscience has indeed been a locus for significant scientific discovery over the last 10 years, but it is hardly a booming sector of the economy. The rise of Applied Mathematics is even more confusing. Applied Mathematics was first founded at Harvard to allow, among other things, students to study computer science before there was a Computer Science department. As Harvard’s Computer Science department expanded, we should have expected Applied Math to decline in influence. Instead, these two fields have exploded in popularity, with the number of concentrators doubling since 2015.
A significant reason for the rise of Neuroscience and Applied Mathematics may be the fungibility of each field with more difficult alternatives. Neuroscience, for instance, is a common path for pre-medical school students, both at Harvard and elsewhere. Unlike other pre-med fields, however, Harvard’s Neuroscience department imposes less onerous requirements on its students. While a student may take the more difficult “quantitative” track, the main track of Harvard’s Neuroscience department, “neurobiology,” does not require a single course that demands more than 3-5 hours of work per week, according to a plurality of students.1 By comparison, Harvard’s two other main tracks of biology, “Molecular and Cellular Biology” (MCB) and “Human Development and Regenerative Biology” (HDRB) require multiple courses that demand at least an average of 6-8 hours of work per week.2
Distributional “Physical Sciences” requirements in traditional biology fields also make them more restrictive and demanding than Neuroscience. For both the traditional biology tracks at Harvard, physical science requirements include courses in sub-fields of chemistry and physics, which are often extremely intensive.3 On the other hand, students can earn Neuroscience degrees at Harvard without taking a single course in chemistry or physics. While these differences may seem small, at a school with such a focus on optimization and graduate school admissions, even a minor increase in GPA can change student behavior significantly. Indeed, the relatively easy pre-med pathway offered by neuroscience has been openly cited by students when asked why they chose it as their concentration.
The popularity of Applied Mathematics is even more clearly a function of relative grade inflation. Students do not just concentrate in Applied Mathematics; they concentrate in Applied Mathematics with a different application concentration. This means that one can get a Computer Science degree at Harvard, or at least a degree that says “Computer Science,” without taking the ordinarily required Computer Science courses. The idea of an Applied Mathematics/Computer Science might seem like a befuddling notion at Harvard. Harvard’s Computer Science department offers plenty of math options for concentrators. The department requires students to take multiple “theoretical computer science” classes, that is, classes in applied mathematics, to graduate with a Computer Science degree.4 These are usually the most difficult courses a Computer Science concentrator is required to take. Indeed, Harvard’s Computer Science department is considered unusually focused on mathematics and theory. It is hard to imagine students looking at Harvard’s Computer Science department and viewing the department as not focused enough on applying mathematics!
Unintuitively, Applied Math is likely popular because it allows students to avoid complex mathematics. Harvard requires Computer Science concentrators to take at least two of Computer Science 1200, 1210, and 1240, three of the most challenging math-focused courses required for concentrators.5 To be considered for honors, a student must take Computer Science 1240, which requires an average of ~15 hours per week. If this weren’t enough, all Computer Science majors must take Statistics 110, a probabilistic mathematics course that requires an average of approximately 15 hours per week.
Applied Mathematics/Computer Science, however, has very few requirements.6 While Applied Mathematics/Computer Science concentrators must take CS 1200 and “2 more core courses drawn from the 1200s, 1300s, 1500s, 1610, 1750, or the 1800s,” this is far easier than the policy of the Computer Science department. Applied Mathematics students also need not take Statistics 110, removing one of the most difficult math course requirements from the Computer Science degree. Thus, Applied Mathematics fills a unique niche for those who want a degree in computer science without taking many of the most challenging courses.
This is not to say that getting a degree in Neuroscience or Applied Mathematics is easy. Indeed, both of these concentrations are significantly more difficult than nearly all humanities concentrations. But more important than the student experience is the fact that grade inflation has redirected many students to undergraduate concentrations that are relatively easier. As a result, when administrators look at what departments should receive greater funding or resources, they see departments that have been chosen as a direct result of grade inflation. Administrators looking for whom to fund may naturally focus on these popular fields, because greater student enrollment means research in these areas is more in need of resources than, say, smaller biology departments. Thus, STEM grade inflation can cause highly distortionary results, where grade-inflated departments receive funding rewards because of their grade inflation.
Of course, academic decisions are not this simple. Both funding cuts and increases are often more functions of donor interest, institutional competitiveness, and academic trends than anything as rational as demonstrated student interest. Nevertheless, there is some evidence that student interest has some impact on academic decision-making. Harvard’s largest donations over the past 10 years, for instance, have been focused on a small number of fields – Computer Science, Neuroscience, and the broader School of Engineering and Applied Sciences – all of which saw significant increases in student interest prior to outside investment. At the end of the day, big donors want their names on big departments. Meanwhile, the humanities, facing a rapid decline in student interest, have seen a significant decline in funding. Student interest plays a significant role in determining which sciences are funded and which are ignored.
While student interest may seem like a heuristic for determining which areas of study should receive more funding, the metric becomes incoherent when there is rampant grade inflation. In an ideal world, students pick fields that are academically exciting or have high economic returns. These are fine measurements of how innovative a field is or how economically beneficial its research, both of which are important factors in deciding what departments ought to be funded. But when students choose a field not because of these value-tracking traits but rather because it will improve their GPA, funding will be misallocated to fields that do not need that money and talent. It is hardly reasonable to say, as Harvard’s behavior implies, that money ought to be spent on Applied Mathematics or Neuroscience research rather than Computer Science or Biology research because of those departments’ grading practices.
Where research is most important and difficult, the university will as a result fail to effectively innovate. Universities are already failing to keep up with research in the most important, innovative fields. In artificial intelligence, perhaps the central research program of our day, American universities are becoming peripheral, producing few of the basic research papers that order modern AI research.7 In biology, too, universities are losing importance, with biotech industry research reaching an all-time high of 35% of basic research output in 2022. Indeed, two of the three laureates of the 2025 Nobel Prize in Medicine, Mary Brunkow and Fred Ramsdell, had never worked at a university, winning instead for their work in private industry.
To maintain Harvard’s role in American research, we must recognize that grade inflation will not just lead to poor repute but a dislocation of Harvard from its research mission. Harvard’s current battle against grade inflation seems more focused on making easy courses harder than in drawing resources and talent to the most challenging and productive areas of study.
To effectively curb inter-departmental grade inflation, Harvard’s administrators must enact policies that are targeted and disruptive. Such policies may seem overly punitive, particularly for those researchers currently benefitting from the relative ease of their departments. But Harvard’s administration cannot take its usual laissez-faire approach to our problems. In Harvard’s recent conflict with the Trump Administration, our most important defense was our research. At a time when the status of elite universities is in question, Harvard cannot afford to lose that defense.
If we fail to stem inter-departmental grade inflation, Harvard will no longer be able to contribute to the most fruitful scientific problems of our time. It is difficult to imagine why a donor, or a government, would view an institution so dislocated from mainstream science as a proper recipient of funding.
In general, Harvard’s grade inflation debate has become repetitive. Traditionally, the debate focuses on fields in the humanities and social sciences which require little work, and even less (if any) good work. According to this view, easy grading causes students to learn less, or makes their degrees less economically useful, or demotivates them. But these concerns hardly threaten the fundamental mission of Harvard; at worst, Harvard becomes a little less trustworthy to employers.
But this story hides a more pernicious form of grade inflation. Because getting into the right medical school or getting the right big tech job is just as easy for someone concentrating in an inflated as an uninflated field, these differences have fundamentally altered what our school considers a significant research question and what our students learn. If we continue on this path, Harvard will lose its dominance as a research institution, and students will no longer learn the skills they need to excel in the fields to which they aspire. Thus, what we need is not more initiatives or task forces on grade inflation in Dance Media or Education Studies. What we need is fair standards across STEM concentrations.
All statements about the number of hours each course requires are supported by the responses on the Harvard Q report when asked about the most recent iteration of this course.
Our statements about the requirements of the Neurobiology degree at Harvard are taken from the concentration’s website.
MCB 60 for those taking MCB and SCRB 10 and MCB 60 for those taking HDRB require 6-8 hours per week according to a plurality of students; both are requirements for their respective fields.
The easiest Physics courses which MCB and HDRB students can take to fulfill distributional requirements demand 6-8 hours per week according to a plurality of students; the easiest Organic Chemistry options average out at about 9-10 hours per week.
Information on Computer Science requirements is taken from the concentration’s website.
Indeed, all three courses require work well-above 8 hours per week of most students.
Information on general Applied Mathematics requirements comes from the concentration’s website here. Information on Applied Mathematics/Computer Science requirements comes from the concentration application’s website here.
Ahmed, Nur, et al. “The growing influence of industry in AI Research.” Science, vol. 379, no. 6635, 3 Mar. 2023, pp. 884–886, https://doi.org/10.1126/science.ade2420.