Monday, December 10, 2018

IITK Revival: Must start from Computer Science

A few decades ago, after the JEE results, I went to IIT Delhi for Counseling, I just put up one option, "IIT Kanpur Computer Science." The good professor on the stage warned me this may not be enough. The previous year's closing rank for IITK CS was 26. And I added my second choice, "IIT Delhi Mechanical." Thankfully, I got my first choice.The top 100 ranks, in those days, would go for IIT Kanpur and overall about 50 percent of top 100 students would be in Kanpur. This continued for many years, but started declining about 20 years ago, and for the last three years, we have had 0 students in top 100.

While there are many reasons for top JEE rankers to not choose IIT Kanpur and the situation is not likely to change overnight, we must do something to counter this perception that IIT Kanpur is in some sense worse that IITB, IITD, and now even IITM, and the rate at which the closing ranks of IITK and KGP are closing in, may be tomorrow's perception will be that we are the worst old IIT. It is nobody's argument that a 100 ranker is far better in academic preparation than a 200-rank student. After all JEE is a lottery among good students. So all the students you are getting are good. But the perception still makes a difference. Having a bunch of 1000 students most of whom preferred some other IIT and are sad (at least in the beginning days) that they are stuck in IITK is not the best environment to start converting these young adults to great engineers.

Based on current perceptions in the admission market, it would be difficult to convince students (and parents) to prefer Kanpur over Bombay/Delhi for the same discipline. But we must note that some students do prefer Kanpur CS over Electrical at other IITs, and most students will prefer Kanpur CS over the next popular program (Mechanical or Maths+Computing) at other IITs. So the preference of other IITs is not so much that students will take any discipline there over any discipline at Kanpur, though if the choice is between the same discipline, they will prefer other IITs. We should be able to make use of this dynamics.

Why do people prefer certain disciplines - not because of their interest, but because of their perception of job market. Even the choice of IIT is driven in part by perception of their placement. It is assumed that placement depends on discipline and IIT and hence you can go to a "lower" IIT if you are getting a "higher" discipline. (I really feel sad to use terms like lower/higher, since they are more based on ignorance and not on passion or an understanding of career issues.) What if we could tell people that if they were to join IITK, they would have wide variety of jobs available, not just in their core disciplines but also in CS related fields, and I am not talking of IT service industry but "higher" end jobs. Even in their own disciplines, they would have faster career progression since they will have skills that are in demand. Yes, that means, every student in IITK, irrespective of their discipline, will be able to do a course on Data Structures, a course on AI or Machine Learning, a course on Robotics or automation, and may be a couple of others, as electives.

We don't have to change our curriculum for this, since it is already very flexible, it already has a large number of elective slots, and the only thing that is needed is that such courses be offered in all semesters, may be multiple sections, and be available for enrollment for all students. While some of these courses could be offered by departments other than CSE, doing this will put a significant load on CSE department. And that is where the problem is. Why should CSE department take up extra teaching load?

It is clear that if we can allow all students to do certain set of courses (as electives) many of which will have to be offered by CSE department, we could improve the closing ranks of other departments. This won't help CSE department, but they would be helping the Institute if they can manage this. And from my discussion with KGP alumni and faculty, I see that they do have this kind of flexibility for their students and this is helping them.

I believe this is possible through a multi-pronged approach. We can make use of technology and offer some courses in flipped classroom mode. We can make use of guest faculty. We should be able to hire some more regular faculty. We can remove some electives with just 2-4 students and ask the faculty to teach courses with some minimum student strength. We can co-teach certain courses with faculty from other departments. There can be incentives to individual faculty members who are not too busy with research to take extra teaching load. There can be incentives for the department as a whole. I really believe that where there is a will, there is a way.


  1. Hello Prof. Sanghi,

    I just graduated from CSE IITK this summer, and I felt a huge shortage of faculty in the fields mentioned above.

    I wished to work on some idea, but I could find a faculty with similar interests who could properly guide me through, so I had to take upon that idea as multiple course projects.

    I am not sure how difficult it is to hire new faculty, but compared to universities abroad. IITK seems to have a severe shortage of faculty at least in AI, ML, NLP and Computer Vision. And as you said offering more courses would significantly impact research.

    1. Can we at least start with Data Structures to everyone (who wants to take). Can we allow more students in the other standard courses. There was a time when the average teaching load in CSE was significantly higher than the Institute average. Not any more. It is less. For the most popular department, to teach less than the average teaching load of the Institute should not be acceptable. In fact, popular departments should be willing to have an average teaching load somewhat higher.

      As far as impacting research is concerned, I am making two statements. One, no one should routinely teach a course with 5-10 students. If you didn't want to teach, you are in the wrong place. If teaching one course a semester with 50-60 students affect your research, so be it. You are supposed to balance between teaching and research. And remember, every top research university in the world expects more teaching than this. Second, in any department in India, you would have people who have decided that they want to reduce their research commitment and are willing to compensate that with extra teaching. Again, lots of universities in the world are having research-track and teaching-track in the faculty where the difference is actually not much. Instead of 3 courses a year for research track, 4-courses a year are taught by teaching track. In government places in India, if someone decides to do less research, they just do less and not compensate it by more teaching. I am sure there are people in our institutions who despite being a permanent government employee wouldn't mind teaching extra and that would not impact their research since they do less of that anyway.