What are four years for?

I remember seeing this tweet in my feed and being reminded of a long-standing frustration with aspects of the education system in India. While perhaps guilty of some oversimplification, the post discusses an evaluation practice that seems to be fairly redundant yet widespread - having to paste printed snippets of code into a “lab notebook” for an assignment.

It is incredibly confusing how so many engineering students spend four years that amount to very little learning - or at the very least, to learning achievable in far less time. You could ask a graduating student how pertinent their university curriculum was to where they ended up, and seven times out of ten you would be disappointed by their answer. Perhaps you would further be told that other aspects of university - such as clubs, student-run teams - were of greater consequence to their motivations and learnings. It is therefore worth discussing the supposed value this curriculum brings - specifically because universities shape their academic systems around the assumption that students derive most of their educational value from coursework and examination cycles. Ideally, this expectation would be a two-way street.

Plenty of students enter university with the sole purpose of emerging with a degree, and it would be unfair to blame an academic system for our own unwillingness to engage with it. But is this truly something that can be fully ascribed to students?

Many students in India enrol in specific engineering disciplines for reasons other than their own interest. This could be because they are attracted to fields perceived to set them up for financial upliftment - until recently this was computer science, although I am uncertain how the field snuck into being known as an engineering department. It could also simply be from a desire for the privilege (both academically, and from a social standpoint) that comes from associating with a “higher-tier” university. Come the annual admissions season, there is no dearth of questions along the lines of “which do you think is better: branch X at university A, or branch Y at university B?”.

This is not to say that these justifications are completely invalid. With appropriate context, it might make sense to presume that the opportunities and quality of life in some highly regarded universities are worth the trade-off. I instead say this to acknowledge some of the circumstances under which seventeen-year-old students - or, in many cases, their parents - make incredibly consequential decisions about their academics.

Unlike (some) liberal arts universities in India and most universities in the West, students do not have the luxury of choosing a major they might be interested in a year into their programme. We expect students to choose disciplines unusually early, under circumstances where their choice is often only weakly related to academic interest, and then design the subsequent four years as though that initial choice revealed something profound about what they ought to learn.

This is surprising: a system that cannot reasonably allow every seventeen-year-old to explore before choosing a discipline should, if anything, compensate by making exploration considerably easier afterwards. This could mean allowing students to take substantially more coursework outside their departments, pursue research unrelated to their degree, substitute sufficiently rigorous external courses for internal ones, or reduce their course load when they are working on something worthwhile elsewhere.

If a student spends a semester doing serious research at another institution, builds something substantial with a student team, or takes a course taught considerably better elsewhere, it is unclear why a university's instinct should be to ask how this fits around their coursework rather than how their coursework can fit around it. This obviously becomes difficult at scale, and there are good reasons universities impose structure. Complete freedom would probably work wonderfully for a small number of unusually motivated students and terribly for many others. However, it is strange to see the opposite system in place: an assumption that every student benefits from approximately the same amount and sequence of structure, and then treating deviations from it as concessions that need to be justified.

If this were not enough, academic systems within universities continue to pedestalize (for lack of a better word) performative learning. By this I mean that universities reward visible demonstrations of learning rather than encouraging real learning itself. The vast majority of students study to score well in exams, which is particularly harmful since (1) this knowledge tends to disappear immediately afterwards, and (2) selective consumption means you never really bother to “learn” what you are studying.

For instance: looking at previous mid-sem exams, or preparing for a set of “likely” questions is often sufficient to score well - not that I bothered to do even this. A part of this is of course because students can be inherently lazy. However - it might not be helpful to have three sets of exams, quizzes notwithstanding, in the span of a four-month semester. There are extremely simple fixes to this, one of which is to instead conduct open book exams and frame questions accordingly.

It is, however, impractical to assign blame entirely on universities and faculty for this. The system would be far better if it accounted for imperfections: not always having the most qualified or enthusiastic faculty (as is the unfortunate case in many lesser-known universities), or not always being able to fund and support students’ ventures and interests.

An exhaustive laundry list of complaints would also lament the lack of genuinely supportive research environments for undergraduates. I will limit my written disappointment to one particular problem: the desperation for research output volume among faculty at many universities, and the gamified, incentive-driven approach towards research that undergraduates consequently encounter.

Universities want undergraduate students to participate in research, while simultaneously creating incentives that make much of this participation less meaningful. Faculty are pushed towards greater volumes of research output, while students are expected to complete course projects that resemble miniature research papers. It is entirely possible for an undergraduate to end up working on four or five supposed “research projects” in a semester, each expecting some combination of a literature review, implementation, results and, in some cases, novelty.

This is a rather odd way to introduce someone to research. The literature students are expected to find novelty in is often the product of graduate students and academics who have spent years narrowing down increasingly specific problems. Expecting an undergraduate to understand enough of this literature to identify something meaningfully missing and investigate it within four months is already optimistic; expecting them to do this several times simultaneously is plainly unrealistic. This becomes worse when these projects exist alongside exams, assignments, placement preparation and everything else expected of a student during a semester.

Faculty are placed in a similarly strange position. Meaningfully supervising research takes time, and there is only so much attention a professor can give to several classes of students while managing their own teaching, research and administrative responsibilities. When research becomes a mandatory and graded component for everyone, there is consequently very little capacity to monitor whether genuine work is happening throughout the semester.

The eventual standard has to become more forgiving, not necessarily because either the professor or student is indifferent to good research, but because the scale of what is being asked makes anything else difficult. The unfortunate result is that we can expose more students to something called research while leaving them with a worse understanding of what being a researcher actually means.

This has only worsened with the age of widespread AI use. It requires little effort to conjure up (arguably meaningless) drafts of research problems, and even less effort to complete assignments or delegate academic tasks. LLMs are no longer entities we can laugh at for being unable to solve a simple math problem or to come up with efficient code. Academia will not be able to fight this overhaul: apart from the difficulty of reliably detecting AI use, it would be impractical to disallow tools that fit into almost every use case across the world.

There has perhaps never been a more pressing time for universities to design their academic systems around these considerations. The engineers graduating four years from now will likely work in a very different world, one that requires substantially different expertise from what they are currently evaluated on.

Today, advancements in AI systems make having expertise particularly useful: people who understand their fields well are better placed to ask the right questions, evaluate what models produce, and make useful decisions about what can and cannot be delegated to them. The solution, then, is decidedly not to limit LLM use. What must change is what students are tested on, and the approach with which they are tested.

None of these concerns are new. Universities have experimented with open-book exams, flexible curricula and project-based courses. Indian education policy has, for years, called for less rote learning, greater curricular flexibility, interdisciplinary education, more continuous forms of assessment and stronger research environments. It is therefore remarkable that the everyday manifestations of the system these reforms seek to replace have remained so persistent. My understanding is that this is because of existing incentive structures within universities.

It would make no sense to generalize all universities into a single category of actors. We have the highly selective premier institutions in IITs, NITs, BITS and IIIT-H. The problems at an IIT are obviously not identical to those at universities like MIT, Thapar, VIT, or one of the hundreds of engineering colleges most of us have scarcely heard of.

There is nevertheless a useful distinction between the small set of highly selective institutions and the much larger collection of private and public engineering colleges outside this group: the former generally have better faculty, research environments and academic flexibility. But even among relatively good institutions, many of the underlying incentives that make academic reform difficult remain.

Neither group necessarily has strong incentives to implement changes in this direction. After all, the premier institutions must bear a cost to experimentation. IITs in particular occupy an enormous amount of public scrutiny, and strive to maintain their reputation as a safe path to academic and professional success.

Given that students go through an incredibly competitive process to earn admission, there is some merit to these universities wanting to stand by a method that seems to continue producing graduates sought after by employers. The very outcomes that make these institutions desirable then obscure the question of how many of those outcomes were produced by the institution itself, rather than by selecting an unusually capable group of students in the first place.

But it is also not organizationally feasible, perhaps, to expect an efficient, flawless series of bureaucratic approvals for new academic policies at universities that are constantly under the hammer. Curricula have to work across departments, faculty have to be available to teach them, credits and prerequisites have to fit together, evaluation has to remain reasonably fair; and experiments that fail affect hundreds or thousands of students.

Some institutional inertia is therefore probably desirable. The unfortunate consequence is that the same machinery which prevents badly considered changes also makes sensible experimentation slow, particularly in institutions where almost every consequential decision attracts scrutiny.

We must look at a slightly different question when we consider the second group (of largely privately owned) universities. It would not be remiss to say that their direct incentive is to maintain admissions and financial viability. Students are often unable to easily distinguish between universities that fit their goals and those that do not.

The easiest way, then, to signal quality to the public is through measurable outcomes - things like placement statistics, rankings, and median salaries. The things easiest to advertise are not necessarily the same as those that would improve education. It would be irrational to give this up in pursuit of an alternate academic system that people might look at suspiciously, particularly given that this would be a resource-intensive effort in itself.

It is unsurprising that despite a desire for change, existing incentive structures lock universities into operating in the same manner. Students want recognizable degrees and good placements; parents understandably want some assurance that an expensive education will result in financial security; universities want predictable academic systems and outcomes they can advertise; faculty need forms of assessment that remain feasible with large classes; recruiters want grades and other cheap ways to filter enormous applicant pools. Nobody has to particularly like the resulting system for each actor to have a reason to preserve some part of it.

With the advent of artificial intelligence, all of this stands to change over the coming years - which is perhaps why I believe this discussion is a lot more relevant today than it would have been five years ago. These thoughts do not mean I believe in an immediate overhaul of the system we have followed for decades; we are bound by constraints wholly different from universities that enjoy a more privileged circumstance. But acknowledging these constraints must not equate to an indifference to the problems that exist.

We must be willing to ask what a requirement is actually accomplishing before imposing it on every student. If an examination is structured to only reward retaining information for two days, why do we not change the way we frame questions? If a compulsory assignment can be completed by an LLM without the student understanding its contents, what did we intend the assignment to teach? If five mandatory research projects produce less meaningful research than one voluntary project, why do we push for five?

It is easy to forget that four years is a long time - we tend to view engineering degrees as pockets of time that take in students to toss them out as employable graduates. The value of a university education should be that students have enough time and freedom to become interested in things they did not know existed when they arrived. If this doesn’t happen, then perhaps it is time to ask more questions of our academic systems.

← Writing