How is hiring done in IITs/NITs/etc.?

This is an important question that members of the general public and most prospective applicants simply have no clue of. It is important to present some facts. Hiring in government institutes is a very nuanced and complicated process, which does not lead to simple explanations or solutions. Governments have their unique way of working.

Note: A knowledge of data science is required to understand this article. Otherwise, appreciating it is going to be very difficult.

First, consider the old IITs.

1. The professorial committee of the department shortlists the candidates. Roughly 3-4 candidates are selected and most of them end up getting hired.

2. The substantive criteria are: (1) Having papers in the "best places", (2) Good reco. letters (written by distinguished people) and (3) A positive interaction with the department (well-received deptt. seminar and one-to-one interaction with senior or mid-career faculty, roughly 8 hours). There are very few candidates in the "acceptable" cluster -- they are clearly far ahead of the rest. lowest(shortlisted) >> highest(non-shortlisted). This is why the hiring process in old IITs is relatively more seamless as opposed to other institutes. There is little to complain given that the shortlisted candidates are in a different league. Many criteria such as pedigree, age, GPA, etc., can be relaxed. This is because if someone has papers in the very best of places, delivered a great seminar and is coming with strong recommendation letters (very important), the candidate can be considered to be "safe". He/she is unlikely to have major problems and should be able to do well in the IIT system with a reasonably high probability.

3. The selection committee's job is thus easy. They still need to do a few basic sanity checks, which is done and the candidates most often pass them with flying colors. In fact, if the committee starts asking technical questions, the candidate can be sure that his interview has gone south -- the committee doesn't fully believe that he wrote those excellent papers and understands the basics well. Hence, some scrutiny is underway. Ideally, the committee should only be discussing short and long-term research plans.

New IITs, NITs and comparable institutions:

1. In this case, such kind of a dominating set does not typically exist. In fact, you have a continuum. The quality of clustering (silhouette score) is quite poor.

2. There is thus a need to interview more candidates and do a thorough screening: 1 hour per candidate. Online interviews, in this case, don't work very well. An offline interview needs to go into all aspects: research ability, teaching ability, knowledge of basic concepts, ability to connect with people, ability to run a research group and get funding, everything.

3. Conducting the interviews is thus quite tough and time-taking. Consequently, it is very hard to assemble a panel of experts, get them to a location that may not be close to an airport, conduct a 1 or 2-day interview and pay them at sarkaari rates (roughly 4k per day). You need a team of people who are genuinely interested in societal service. There is no other serious incentive. Often they are alumni of the institute who feel for it, personal friends of the Director who are obliged to agree, or just genuinely nice people who can't say NO. The Dean Faculty and Director struggle to put a committee together, find dates and orchestrate the entire process.

4. Given that there is no financial or other kind of incentive, NITs and IITs have to work very hard to assemble a committee: bribe committee members with good-quality sambhar (perfect the balance between tamarind, herbs and heeng), make the best-quality green tea/black coffee or arrange a trip to the local temple at 5 AM in the morning [priority-access]. The guest house cook often decides the quality of the selection committee !!! Maybe just humor the committee members with good anecdotes and jokes. That works.

5. Given the tremendous difficulty, conducting an interview even once every year or two is very tough. Hence, all efforts have to be made to ensure that the 10-20 candidates that are interviewed, are the best in the pool. Also, the board needs to approve the full list of selected candidates for all departments before another interview round commences -- this in itself creates a delay of 6 months (at least). Expect multiple court cases, RTIs and grievances. This further delays the process.

6. Now, how do you find the "best"? A separating hyperplane needs to be found that minimizes misclassification. Assume a hypothetical ground truth where every candidate is interviewed by the committee. It is a case of unsupervised learning, which can be made semi-supervised if you consider past experience and some knowledge of the prior. You are clearly not expecting Nobel laureates or Field medalists. In fact, consider yourself very lucky if a CS candidate can write the code to find the third-largest number in a list in O(N) time, or an EC candidate knows the difference between a fermion and a boson, or the natural oscillating frequency of an RLC circuit. In fact in CS, it is not uncommon to encounter candidates who cannot write code to test whether a number is prime or not, or even find if it is even or odd !!!

7. Let us next look at the available knobs and their pros and cons:

a. Class 12 marks in PCM or total: a standardized metric that arises out of large-scale testing. Students are close to the age of majority. On the flip side, boards have a lot of variation and the grading may not be up to the mark. What if you join a board that has arbitrarily decided to award 71% marks to the topper, even if the topper is Einstein? That said, the Damodar Acharya committee did find a good correlation between board marks and known metrics of success in an engineering career. What is the guarantee that a board that awards 71% to its topper has created a topper who can be employed as a faculty? Does the board grade engineering papers the same way the British used to grade English papers or did the topper genuinely not know the answers to 29% of the questions? Many questions here, no straight answers.

b. UG pr PG college NIRF rank: Which one do you take -- current rank or the college's rank when the candidate had joined? What if the candidate joined before NIRF was created? What if the candidate reported sick on the day of JEE/GATE or could not afford coaching? On the flip side, NIRF has some value even though there is no dearth of articles that point out it flaws. After all, someone from a better institute has had the benefit of rigorous training for 4-6 years, which is very difficult to supplant or compensate for. Also, the candidate must have done well to get into the top-ranked institution in the first place. This cannot be denied. This is a game of probabilities. Higher the "perceived rank" (quantified by NIRF to some extent, some may say quite erroneously) of the university, greater the probability of finding a better candidate. We will discuss the legality of choosing high-NIRF institutions towards the end of this article by courting a famous HC judgement.

c. UG or PG marks: Here also, there is a wide variation in colleges. Would you prefer 7.5 from an IIT or 9.5 from a private deemed university, where a blank paper fetches 8.0?

d. Additional qualifications such as GATE score, post-docs, patents: These are not standardized metrics. Most candidates may not have written GATE or done a post-doc.

e. Papers: In EE/CS almost all these papers are either in AI or device physics. It is not uncommon to find candidates with 50 papers applying for a Grade-I Asstt. Prof. position. All these papers are in unknown places. Nothing can be said about them. The SNR (signal-to-noise) ratio is so low that using papers as a sole metric cannot simply be justified in any way, scientifically or legally. The paper lists across candidates are simply incomparable. AI indeed weaves its magic here. We have a partial order and no semblance of a total order. The count of the number of SCI-indexed papers is also a very weak signal given that SCI indexes a lot of conferences and journals, whose quality can radically vary. The quality is all over the spectrum and having 10, 20 or 30 papers means absolutely nothing. It is basically AI or ChatGPT magic.

8. Hence, the only option in front of institutions in this category is as follows:

a. Find the maximum number of candidates that can be interviewed after consulting with the selection committee. Let this number be N. The institute will like to maximize N and the selection committee will like to minimize it. Let N' <= N be the number of "good candidates" in this set of N (learnt post-facto). The aim is to maximize the expected value of N'.

b. Create shortlisting criteria that is (as far as possible) consistent with yesteryear's criteria. If it was acceptable last year, it will be acceptable this year. If the number of candidates is less than or greater than N, then modify the criteria. The problems thus begin ....

c. [Principle 1:] Never violate any law of the country, case law, or the acts and statutes of the institute.

d. [Principle 2:] Do not create unreasonable criteria: a GPA cut-off equal to 7.7234 (should be either 7.5 or 8.0). Declare all conflicts of interest even if your neighbor's girlfriend's third cousin is applying. The criteria cannot be 3.4 * #sci-papers + 4.2 * #awards + 2.9 * #GPA. It "appears" too arbitrary and will not pass legal scrutiny. You need to be fair and also appear to be fair. Follow the Occam's razor and have simple criteria.

e. [Principle 3:] Tune the knobs (in (7)) to ensure that there are N candidates. Maximize the precision. Use data from previous years and the accumulated experience of the community to create some sort of a semi-supervised learning system. Here, the past experience of that particular institution counts more than anything else. The hiring department only knows what formula works the best for it. The Head has a lot of role to play here in consultation with the Dean Faculty.

9. Do whatever you want, finally it is the Fano's inequality in information theory that will establish the efficacy of the shortlisting criteria. If there is some degree of uncertainty in the knobs, of which there is plenty, then there is a hard bound on how good the separating hyperplane can be. Also the candidates may not be linearly separable especially when you consider nebulous metrics such as papers. 1 + 1 is not 2 here. You are bound to miss out on some good candidates and shortlist some not-so-good ones. A few of the non-shortlisted candidates will cry hoarse, file RTIs, go to court, take to social media, and pull every string that is out there. Nothing can be done. This is where the Shannon entropy (lack of information) hits you hard and mistakes will happen.

10. Fighting court cases and answering RTIs/grievances is a template of our times. All academic administrators can pretty much do it in their sleep (metaphorically so). Thankfully, courts understand all these practical limitations of institutes and penalize them only when genuine malfeasance can be established beyond reasonable doubt.

Consider the following case. Can a government agency funded by taxpayer money or a PSU decide to conduct campus interviews only in the top IITs? What if there is a good student in a rank-100 institute? Wouldn't this violate his constitutional right to equality (article 15) and equal opportunity (article 16.1)?

The court indeed had to grapple with such a question in M. Palanimuthu vs. The Secretary to Government (2015). Let me reproduce a few lines from the judgement by the hon'ble Madras High Court.

We fully endorse the stand taken by the public undertakings that in a present day scenario they have to compete with multinational private companies and other public sector units and get the best talent and one such method which has been time tested is to go to the place (institution) where talent pool is available. The principle appears to be catch them early. We find nothing arbitrary or unreasonable nor does it in any manner infringes or impinges into any of the Constitutional guarantees.

It basically says that PSU companies can only go to top IITs. Basically, in this world, what is fair and what is not fair is a very complex question and there is no clear answer. The view and viewpoint matters. Is it fair for a PSU's senior officers to roam around the country for months and interview thousands of candidates in the name of fairness, and still end up hiring recruits mostly from a few institutes that were a priori known to be good?

What about age? Can you say that you will only hire Asstt. Prof. candidates below 35? Isn't this age discrimination? Go no further. Let us look at another Supreme court judgement (2022).

& Ors vs Union Of India & Ors on 20 May, 2022.

A group challenged the upper age limit of electrical supervisors in West Bengal. This is what the SC said.

This Court further holds that the State Government was competent to fix the upper age limit by 2017 Rules and the same cannot be said to be an arbitrary one. Since the aspect of safety of the general public is involved with the job of supervisors, the fixation of upper age limit imposed by the 2017 rules cannot be said to be unreasonable. The State Government has the power to impose age restriction while laying down the conditions for grant or renewal of licence, certificate and permit by framing rules. In view thereof, this Court holds that the writ petitioners have failed to establish that they have been deprived of their right to earn livelihood except according to procedure established by law.

Those following Biden vs Trump will concur that mental acuity reduces with age. Of course, experience and training have their value. There is a trade-off here and the hiring institute decides the sweet spot or cut-off. All govt. organizations in India do it. As you can see, the SC has upheld it and the SC itself has set an age limit of 32 for its clerks (for instance), and all other positions right up to the chief justice. Private organizations also do it, but do it more surreptitiously. Look at this Livemint story.

https://www.livemint.com/industry/infotech/wipro-invites-job-applications-from-freshers-eligibility-and-other-details-11641093409716.html

What the cutoff or hiring criteria should be is an open question and there is nothing which is absolutely fair -- everything is unfair to certain degrees. Any shortlisting and subsequent hiring decision has to be looked at from the lens of the contemporary times, the judgements of the courts of the era, practical considerations based on known priors (game of probabilities) and the power of agency of the hiring institutes themselves. They after all take the decisions that they believe lies in their best interest and the same is vetted by the selection committee, who are external members and are often there for the sambhar, black coffee and in the pursuit of serving society.

Moral of the story: Take the best possible decision in the interest of the institute given the information that you have and the knowledge the community has accumulated over time.