The Dream Grading Policy
Reimagining evaluation systems that (probably) supports learning
Introduction
Grading is undoubtedly one of the most frustrating experience in the life of a college student. I would also assume that (most) professors do not enjoy the idea of drawing a line between their hard-working students either. Still, like it or not, grading is necessary to evaluate performance. So the real question is: how can we make it better?
In general, there are two popular grading systems: percentage-based and grade-point-based (CGPA). The TAG policy (twisted absolute grading) I will outline assumes the latter system, but it can be easily tweaked to adapt to the former too. So, let’s not worry too much about grading systems and instead focus on grades at the course level.
Problems with RG
Before going into the problems of relative grading (RG), we need to understand why we even have it, i.e., why it is such an attractive option for professors.
It is a no-brainer that students at IIT Bombay are smart; after all, they have cracked some tough examinations. If the evaluation is too basic, then everyone will have near-perfect scores, making it difficult to distinguish between them. RG allows professors to design scarily hard exams so that they can better divide the students. They can then leverage the Central Limit Theorem and automatically determine grade cutoffs by observing deviations from the mean of a Gaussian distribution. However, this system has many negative effects on students, such as:
- Uncertainty of Cutoffs and Resulting Helplessness
- Since the cutoffs are a property of the distribution of all marks and are not known until all marks are finalized, a single person will not know where they lie on the Gaussian. This leads to a feeling of helplessness, as things are not in their control. Had the cutoffs been known beforehand, they could have prepared accordingly for the grades they wanted. But now their only choice is to give their best, which sounds good in theory, but when students are taking eight or nine courses simultaneously, it is simply not possible for most to manage all courses without sacrificing other aspects of life that IITB itself has to offer. And even after giving their best, good results are not guaranteed (a failure of effort versus expectations), and this can cause students to trigger imposter syndrome leaving students feeling like failures despite their hard work as in the end, they may end up with neither good grades nor other skills, for which they simply didn’t had enough time.
- Competitive Hostility Over Collaboration
- Another problem is the mindset RG creates: instead of just studying harder, students realize they can also “get ahead” by making sure others don’t. That means hoarding resources, refusing to help classmates, and generally creating a hostile environment; sometimes done even out of fear that the success of others could lead to their own failure.
- Misplaced Focus on Grades Over Learning
- Notice how much emphasis is placed on discussing grades, even though they are not within students’ control. Shouldn’t the real focus be on learning? Instead, students become so absorbed in competing for grades that they forget the real purpose of being here. College should be a time for exploration and self-discovery. So, evaluation should put learning at front and center, with grades as a side effect.
The Solution: TAG
Well, let’s see how TAG can resolve the problems of RG while still preserving some of its benefits. Because we do not know the Gaussian distribution beforehand, the scheme fails at providing exact borders, so instead we underestimate these borders. The key ingredients here are leniency and flexibility, achieved by grading students not out of \(100\) marks but out of \(100 + x\) marks.
For example, consider an application-type course with a reasonable amount of programming. The evaluation criteria might be as follows:
- \(25\) marks for exercises (one question per lecture)
- \(45\) marks for three assignments (in groups of one or two)
- \(30\) marks for project (same group as assignment)
- \(20\) marks for endterm exam
- also \(+5\) bonus marks for best assignments and project
These add up to \(120\) marks in total, but grading still works like the usual absolute scheme: \(90+\) for an AA, \(80+\) for an AB, \(70+\) for a BB, and so on.
For people unaware of IITB’s grading system, students are awarded a two-letter grade for completing each academic course based on their performance. Each letter grade is converted into a numerical equivalent on a 0–10 scale, and the points of all courses are then weighted according to their credits to calculate the final CGPA.
Letter Grade | Numerical Equivalent | TAG Cutoff |
---|---|---|
AA | \(10\) | \(\in(90,120]\) |
AB | \(9\) | \(\in(80,90]\) |
BB | \(8\) | \(\in(70,80]\) |
BC | \(7\) | \(\in(60,70]\) |
CC | \(6\) | \(\in(50,60]\) |
CD | \(5\) | \(\in(40,50]\) |
DD | \(4\) | \(\in(30,40]\) |
FR | \(0\) | \(\in[0,30]\) |
How TAG Solves Problems of RG
- The biggest win here is these \(20\) extra marks, inspired by Prof. Shivaram’s Autumn 2021 iteration of CS747, which serve as leverage for students:
- If they feel they can achieve the grade they want without completing everything, they can pick and choose the evaluations they prefer (for example, skipping exams) and redirect their efforts elsewhere. The lure of bonus marks for best submissions (from CS683) also ensures that students with this intention still submit their best work, as earning them makes it even easier to skip other evaluations.
- If they feel they need a better grade, they can go the extra mile by doing more evaluations, and these additional efforts will directly affect their grades.
- Firstly, students know the cutoffs, so they know what grade they should be aiming for.
- Absolute grading completely eliminates the idea that any grade is limited and hence can bring students to study together.
- Even at \(75\%\) of marks (which is \(90\)), a student can get an AA. And for failing they need to get less than \(25\%\). Combined with the leniency of exercises (to be explained below), even relatively low relative grading cutoffs can be mapped effectively into this system.
Most of these elements are actually drawn from multiple positive experiences I had with different courses at IITB.
More on Grading Elements
“Easy” Exercises
Many courses give marks for attendance, which never made sense to me. Forcing students to attend does not guarantee that they will pay attention in class. But, what might help is the idea of “easy” exercises, which comes from my experience in Prof. Debasattam’s courses, where he would assign “exercise left to the reader” problems. These tasks varied from proving a skipped small step from a proof covered in class or solving an application of learned concepts. They weren’t necessarily hard problems and usually took no more than 15 minutes. So, I think assigning one such problem for each class (due before the next session) would not only save class time by speeding up content delivery but also help students review the previous material before attending the next lecture, avoiding that classic “blank face” moment when the professor asks what was covered previously :).
An “exercise” for you, the reader, is to generalise TAG to percentage-based grading systems :D
“Tough” Assignments
Here, my inspiration comes from Prof. Rajwade’s class. These questions can be very challenging and time-consuming, but this works because assignments are completed in a more comfortable environment than examinations, allowing students to be tested more deeply.
The assignments will include:
- theoretical questions requiring deep understanding and application of the concepts covered,
- programming problems implementing algorithms discussed in class, and
- fun open-ended explorations involving reading and understanding relevant research.
This approach contrasts with some courses (especially theory-heavy ones) that fill assignments with practice exam questions, failing to take advantage of the take-home nature of assignments. I think three such assignments (with 7–8 main problems each) can balance the busyness of student workload with the satisfaction of fulfilling course’s potential.
These well-designed assignments, combined with the project, can be consistently spaced a month apart, providing enough time for exploration. Though these long gaps still weren’t enough in Prof. Mythili’s course and students kept begging for extensions; but I would like to give it another shot :(.
“Easy” Endterm Exam
Let’s be honest: very few people actually like exams, especially when they are just about vomiting out what was mugged up the night before. I was once a big fan of open-book, or cheatsheet-based exams, but I realized that they often required too much effort (such as cramming information onto a single page) and provided little real benefit in solving problems as this just made the exams harder. And, this is also unnecessary as students are already tested through challenging components like assignments, and projects. Instead, exams should simply check whether students have held on to the big-picture of the course and can apply them while being proctored.
Instead of allowing arbitrary notes or cheatsheets, I think the best approach is to provide a uniform “reference sheet” to everyone containing essential formulas and concepts that students might otherwise struggle to recall. I came to appreciate such exams only in my final semester, when I discovered the effectiveness of a simple revision session before the exam, which can even achieve full marks.
Apart from Prof. Rajwade; Prof. Soman and Prof. Joseph John had also integrated this idea in their theory and lab-heavy courses respectively.
“Tough” Project
This is the big one isn’t it, and it can be a creative option too with multitude of ways that it can be done. I look at it as a big assignment where students create a problem themselves and attempt to solve it, either through their own ideas or by using existing research. Naturally, this makes the project the most subjective element of the scheme, but that subjectivity also allows for greater leniency in evaluation.
While a project works well in most courses, it is not strictly necessary. If a project is not included, its weightage can be adjusted by adding another assignment and/or a midterm exam, but there is another interesting alternative of Peer Reviews, which could either take a smol portion of the project marks or be allocated the full 30 marks.
“Peer Reviews” (optional)
The idea is simple: for each assignment, every student reviews two submissions, gaining insight into different writing styles and reasoning. Then their reviews can then be evaluated based on whether they successfully spotted weak arguments/programming mistakes. I encountered this form of evaluation in CS228 (Logic), taught by Prof. Krishna, and I think it could be especially effective in theory-heavy courses.
Peer reviews work nicely as an individual activity. They make sure students look at all the assignment questions instead of just the ones they worked on in the group.
Plagiarism and Punishments
Cheating is probably the ugliest part of evaluation, especially for take-home work. But, it can’t be ignored either, here are my takes on it depending on the type of evaluation:
- Exercises
- The goal of these exercises is to help students understand and keep up with lectures. Since answers are often already available online, these can be treated as essentially free marks, as long as students cite their sources, whether it was the internet/friends, or even AI (in which case the entire chat history can be shared)
- Assignments
- Assignments are designed to be difficult and often have multiple valid approaches, especially in programming. So, discussions with peers can be permitted (must be acknowledged in the submission), but direct copying shouldn’t not allowed. And if anyone uses internet/AI they really need to be upfront about it by citing appropriately.
- Project
- Projects are usually unique, so plagiarism is less of an issue and can be handled on a case-by-case basis if needed. Still, proper citation is required for any work or ideas that are not original.
And honestly, if after all this leniency someone still cheats, they will have to face the strict punishments. At that point, the professor should not feel at fault for not having done enough to prevent the situation, as they already gave students plenty of room to play fair.
Problems with TAG
This scheme does not fully account for students who are unable to participate for extended periods of the semester due to health or personal emergencies. The 20-point cushion may address some cases, but it is not always fair, so some scaling adjustments would be required.
The other issue is the workload, all the teaching assistants would probably be screaming while reading this blog ;). As for big classes, this is a lot of work for them, and adding peer reviews would mean they would have to double-check assignments, which could get overwhelming.
Conclusion
Of course, this scheme is not perfect, and the mark distribution is kinda arbitrary. However, the goal was to present key ideas, which can inspire you to design an evaluation scheme that best fits your own context :)