Search This Blog

Thursday, October 19, 2017

Fare hike by DMRC

First, a caveat. I am just a railfan who wants rail based transportation to do well. I am not an economist or a sociologist or a ... So take everything below with a pinch of salt.

Delhi Metro fares were hiked last week, a second hike after 5 months. But let us remember that this was because there was no hike since 2009, and the Fare Fixation Committee decided that a very large fare increase be done but split into two phases. Naturally, there are people supporting fare hike and others opposing fare hike. Those who support will argue that if passengers want good service, they must be willing to pay for the cost of that service. Those who oppose fare hike point out that society benefits a lot in terms of reduced traffic, pollution, etc., and hence society should have to pay too (through a subsidy by the government from your tax rupees).

Pro-hike people would want to point out that society has already paid and continue to pay somewhat. The land was given cheaper, the right of way, certain tax benefits, and some land specifically for development and earning money were all given by the government. And anti-hike people will say that that was too little. The benefit to society was lot more and continues to be lot more. Everyone would agree that DMRC should try to generate more non-ticket revenue - advertisements, shops on stations, etc.

I see this debate as problematic, since the beneficiaries are either passengers who can be identified and asked to pay for their specific journeys, and society-at-large which means government must subsidize through taxes since individual beneficiaries cannot be identified. The debate would be very different if we can identify other beneficiaries of Metro and ask them to pay.

Are there such identifiable classes of beneficiaries. Indeed, there are, and they were identified even before the first KM of track was laid. Somewhere down the line, they were forgotten or lumped into society-at-large.

First, the property owners. The value of properties near Metro stations have gone up. Also, there was plan of allowing higher FAR to residential buildings near Metro stations. If property owners have benefited from Metro, they should pay for it. The property tax collection should go up, and a part of the increase should be shared with DMRC.

Another category of users who have benefited from Metro are users of private vehicles. While it may sound strange that because of Metro, vehicle owners are able to drive faster, but imagine what would have been the traffic like if an additional 30 lakh people were on the road in a day. So people who are driving are a significant beneficiary of Metro and must be asked to pay. One way to do so would be to levy an appropriate Metro cess on every litre of petrol/diesel sold in NCR and that money shall come to DMRC. (Government may decide to not increase the cess but give the same amount of money from the significant amount of taxes that it already is collecting.)

I am sure one can identify other beneficiaries of Metro and find a way to charge them. If Delhi Metro is paid for by all its beneficiaries instead of mainly passengers, the fare would be less than what is being charged, and DMRC would not be in financial stress.

Of course, I have no idea on how to apportion costs to different beneficiaries and hence I am refraining from giving any numbers.

Wednesday, October 18, 2017

Is Biometric the right way to avoid Proxy Attendance?

Most academics in India seem to believe in compulsory attendance at university level. Unless you have some minimum attendance in the class, you should be punished by either failing the course outright, or in some enlightened institutions, by one grade less. (Of course, then we have places like IIT Kanpur, which leaves the attendance issue to its instructors.)

The logic that is often given is that people have often found a positive correlation between attendance and performance. That Indian kids were never given freedom while they were at school, and hence they are not ready to be responsible adult and thus must continue to be forced to do certain things in their own interest.

I, of course, disagree, and believe that if we will not let them be responsible for their own actions and inactions, they will not learn to be adults. So I don't take attendance in my classes (unless it is the institutional requirement in the university where I am teaching that course and I have no choice).

I believe that there is a positive correlation between attendance and performance. I can give an anecdotal evidence. In the course that I am teaching this semester, if I look at the mid-sem exam answerbooks that have not been collected (because the students were not present in the class), the average is much less than the class average. Also, in the last class before Diwali, I decided to announce the top 10 students of the class in the mid-sem exam and offer them a chocolate each. I found that all 10 were present in the class even on a day when the attendance was only 60 percent. So students whose performance has been good are regularly attending the classes.

However, I am not convinced that the correlation between attendance and performance can be considered as a correlation between attendance and learning. My own gut feeling is that if you attend all the classes, there would be some random stuff that you will be able to recall in the exam (and any exam would have at least a small component of recall type questions). For passing the course, that may be enough, but if the goal is learning, we may not have achieved much by insisting on attendance. (So poorly motivated students would not learn much from lectures, but they will get a few extra marks in the exam to pass the course is my gut feeling. I don't know how to verify this through an experiment.)

I also believe that to the extent there is correlation between attendance and performance, we can advice the students with data and let them decide if they want to really attend the class. Frankly, I have no problems if they can learn Computer Networks from the best faculty members in MIT and Stanford through MOOCs. And if they waste their time and don't learn, they should be prepared for a poor grade, including a Fail grade.

Many faculty members would say that failure has a cost. Not only do they become unpopular for failing many students (and they don't want to be) but also, the class size becomes larger, the infrastructure needs are higher, if a student has to stay on for an extra semester, then hostel accommodation becomes an issue. And hence forcing attendance is ok. (I don't agree that you have to fail a large fraction of the class to get the next batch to become serious. My own failure rates, other than failures for copying where I am very strict, have been pretty much in line with Institute averages, and without any requirement of attendance, I always have had a large attendance in my classes.)

But is taking attendance without costs, and does it really give you correct data. In a large first year class, the immediate question is how do you avoid proxies. If you don't do anything about proxy attendance, you have completely wrong data which is pretty meaningless. So what do you do about proxies.

You obviously don't want to take attendance manually and then enter the attendance into an excel sheet manually. Typical methods include swiping your college identity card in some machine, or having an RFID chip and the attendance machine is really just an RFID card reader, or having WiFi access point figure out whose mobile phones are connected to the AP within the class, etc. All these methods which require the student to own something and carry that with them to the class, have one problem. The student requesting the proxy can easily share that ID card or phone with another student who is going to the class and will be marked present.

So then we come up with the nuclear bomb - the biometric based attendance. This seems to be the seasonal favorite. We use them in most colleges and universities. That avoids proxies (at least for now, there is a news item that people have been able to make a stamp of finger prints which can be used to mark attendance as proxy), but does it bring in any new problems.

While the proxy is avoided, one notices that a student enters the class, marks attendance, and walks out. Has he really attended the class. Are you now going to put cameras everywhere and build software that would recognize people who come late or leave early and mark them absent. Surely, technology can do a lot of things for you, but at what cost.

What if the biometric database is leaked. Are our institutions really well equipped to keep someone's personal biometric data safe.

If we had used one of those non-biometric mechanisms for attendance, and checked for proxy on any random day, and if someone is found guilty, s/he gets a severe punishment, wouldn't that minimize the proxy to the level that the attendance data is reasonably good quality. Do we really need to put every student at a risk of making their biometric data public.

So, why do we use biometric. It does not solve all problems and introduces a huge problem of possible leakage of biometrics.

Well, you ask this question from the students. We have two options. One, we will use a mechanism without bio-metrics, will check for proxy once in a while, and if someone is found guilty, then that student will get an F grade in that course. Two, we will use a bio-metric attendance method. We will never check for proxy or any other problem. But there is a risk of bio-metric data becoming public, and you educate them on what it means for biometric data becoming public.

You can be sure that most students would prefer a non biometric attendance. So you implement that as an administrator. One fine day, the instructor catches a student and tells him after due diligence that he will get an F grade in the course. I can bet that all hell will break lose. This is too tough. Proxy is done everywhere. You are spoiling someone's career for such a trivial thing. May be a warning can be issued or at best some 5% weight could be there for attendance in which this student can get 0, and so on. They will forget that the system was something that they chose. They signed some piece of paper.

Admins know this by experience and do not wish to agitate the entire student body. So they won't ask students for their opinion at all, and just implement the biometric attendance.

I really think that biometric technology is great and should be used when the security and safety is of utmost importance, but to use this for solving small problems like attendance in a classroom (which is not even a problem according to me) can be counter-productive. But administrators want a solution which will lead to more peace with students in the short term.


Thursday, September 21, 2017

Completing 24 years at IITK

Today I complete 24 years as a faculty member of CSE Department at IIT Kanpur. It has been a fascinating journey, and I have learnt a lot. The student interaction has been particularly satisfying. So here are some incidents/events that I recall on this day. I realize that people write these blogs on 20 years or 25 years, and not at 24 years, but I don't want to assume anything about the future, and I feel like writing today.

It is long. So take your call whether you indeed want to read this.

The first thing I remember about the department is how poor it was when I joined. As a graduate student, I had two workstations on my desk, each with 1GB hard disk, both full of network monitoring data. When I was coming to India, I copied that data onto 20 tapes of 60MB each. My advisor told me that it would be useless since I won't have so much storage/compute power to do anything with this data. But I was like, if I can have two workstations as a student here, the premium computer science department in India would at least give a workstation to its faculty. How wrong I was.

When I joined, I was pampered by the department, and I was given the PC with the best specifications. A giant 40 MB disk. Out of this, 20 MB was consumed by Windows, and I was left with 20 MB of diskspace for my files. The Computer Center used to give a disk quota of 10 MB, and the CSE lab gave me yet another 5MB. I pleaded with both CC and CSE to give me 60 MB of space for a few hours. There were two tapes which had the most important data with which I could start working with. I need to copy the entire 60 MB on a disk, and then delete 40MB out of it, and move the other 20 MB to my PC and then I could do some post processing. But for the next six months, I would not get 60MB. I was promised that if CC or CSE buy any new server then before making that server public, I would get access to it for a day. And that took six months.

CSE Department had a total annual hardware budget of Rs. 1 lakh, and an entry level server would cost more than that. So CSE department could never buy a server unless in some year, the administration was kind enough to give us a special grant to do so. Most of the servers and infrastructure in the  lab, including PCs, were bought through projects by faculty members. (Can you imagine today faculty members placing their resources in a common lab in any department. But that was the CSE department culture at that time. And it was this culture that had attracted me to IITK to begin with. The labs are considered personal fiefdoms of the faculty in most departments even at IITK, not to mention the other institutes.)

It was a very tough period. And frankly, CSE departments in other IITs were not this poor. Our budget on a per capita basis was the lowest compared to all other departments in IITK. I had other options, and I thought about them. But the person who mentored me throughout this difficult period was Gautam Barua. But for him, I would have left IITK soon after joining.

The department was a very cohesive unit, and it took care of everyone. For one whole year, before I got married, I only cooked on Saturdays and Sundays. Lunch was in some hostel mess. Dinner was fixed. Every Monday at Sanjeev Kumar's place, every Tuesday at Ajai Jain's place, and other days by rotation among several other faculty members. The social life was absolutely fantastic. So many kids' birthday parties, potlucks, picnics - sometimes with families and sometimes with students, and sometimes both. Today, we are more professional and such get togethers happen within small groups, that too infrequently.

Pankaj Jalote realized that the only way to build the department was to get more money, and the primary source of that money could be industry. So he started something called Industry Affiliates Program (IAP). In those days, any faculty member wanted to initiate anything, they would be encouraged to do so. In IAP, we will invite industry to become our partners by paying a small annual fee. In return, we will inform them of every project, thesis, etc., happening in the department. We will invite them once a year for discussions on various issues, including joint projects and curriculum, etc. After a year, this responsibility was given to me. I was much more direct in asking for funds. I never had shame in asking for money. And we received a large donation from Verifone. It was around Rs. 35 lakhs. Let that sink in. A department with an annual budget of Rs. 1 lakhs suddenly get a check for Rs. 35 lakhs. That was the largest donation that IITK had received till then, and was more than all the donations that IITK had received in the year through Dean of Alumni Affairs office (called DPRG at that time).

And the way it happened was interesting. Verifone decided that they wanted to donate Rs. 1 crore to Indian academia, and to ensure that this had some impact, they also decided that they will not spread this too thin, but divide this money into only three departments. So they wrote to who ever they knew. And thanks to Industry Affiliates Program, we were on their radar. Prof. Phatak from IITB immediately called them, arranged a meeting and flew to Bangalore. I did not have budget to fly, but I called them, sent them a presentation, invited them over to Kanpur, and there was no third department who approached them. They told me that since Prof. Phatak was most pro-active, IITB will get Rs. 50 lakhs, since I was next, we will get Rs. 35 lakhs, and some third department will get Rs. 15 lakhs.

This was really the turning point. We had a "Verifone Lab" in the department for the next few years, which could take care of all our requirements. And within that time frame, thanks for Pankaj's stint in Infosys and knowing Mr. Narayan Murthy, we received a very generous donation from Mr. Murthy.

But before that, other problems had to be faced. We were suddenly told that there is very little money for MTech and PhD fellowships. So we could only admit so many PG students with assistantship. If we wanted to admit more students, they would have to be on "self financing" basis, which means that they won't get any assistantship and their tuition was also higher. But, if we could get industry to sponsor their assistantship, they will continue to pay lower tuition, and the department would get a matching grant for another MTech student. With my credentials as the fund-raiser of the department, I was put on the job. And we got more industry fellowships than all other departments combined. So much so, we did not have to reduce our admissions by even one student, and the Institute said that they did not have enough money to give matching grant.

Another major funding happened by IBM. They wanted to get into the training business which was completely dominated by NIIT and ApTech at that time. So they prepared a curriculum and approached us to develop coursework for them with fairly generous terms, both in terms of initial money, and royalty over the next few years. Normally, we wouldn't accept such an assignment. But the money was so much that it could transform our labs. More than half the faculty was involved in developing course material for IBM. (They started this business, but couldn't succeed, and later used those notes outside India, and still gave us royalty from the money they earned in other countries.) Now, after taking care of all department costs, and Institute overheads, there was still a lot of money left. As per the norms, this was personal money. All of us who had build that course work sat together and decided that we will take a small amount of honorarium and donate the rest to the department. What we donated to the department was more than 6 months' salary for me (I think it was closer to a year). And we did that because our labs and other infrastructure despite Verifone lab was still considered inadequate, and we weren't going to get any money from the Institute.

Let that sink in too. More than half the faculty members donate several months' salaries to build department labs. I don't know if this has ever happened anywhere else in the world. That was the commitment of the faculty towards academics, and I was and still am proud to be part of such a group.

Of course, once we had the gift by Mr. Murthy, there was no looking back. Once we did not have to worry about small moneys, the focus shifted to research. Not that we didn't do research earlier. My group was the first one in the world to have a minimal working implementation of IPv6 for Linux. Unfortunately, we couldn't carry that forward for lack of resources.

We were housed in the limited space on the first floor of Computer Center, since a separate program was created in beginning of 70s. As the department grew, there was very little space for anything. I shared my office with another faculty member for a few years. I don't think any other department had that problem. Very limited space for keeping lab PCs. No space for PhD students. No meeting room. Only one lecture room.

We were overjoyed when we heard that some of Mr. Murthy's donation would be used for a new building that will house CSE Department. Pankaj Jalote was the head, and he involved each one of us in deciding our specifications, interacting with architect, and later interacting with the works department, and also for the interiors. Of course, he wanted this to be a modest building, a modern building but no luxuries. And what came out was one of the best designed in terms of usability of every inch of space. This allowed us to expand PG programs, faculty, new labs, projects, and everything else. PhD students had their own personal spaces now.

 I was extremely lucky that Pravin Bhagwat and Bhaskar Raman decided to join our department. Together, we received a huge Media Lab Asia grant to build the world's longest multi-hop network based on WiFi. This was my first time to work with poverty and rural India. It was so surprising that there were places near Kanpur-Lucknow highway where the nearest public telephone booth (PCO) would be a few KM away. And, of course, the nearest Internet cafe was perhaps 10-20 KM away, or even more. Poor persons were willing to pay double or triple the BSNL charges for making a phone call to their son in Mumbai, since their only other option was to lose half a days' wages by going to that PCO in the other village. We also did research on how poor people consumed Internet by setting up an Internet cafe in a village not too far from IITK campus. This was probably the most exciting times in the department.

Later, me and Deepak Gupta worked with some not-to-be-named agencies on monitoring Internet. This was a self-inflicted disaster. We worked on it for 6 years and at any point in time had a product far superior to what the Indian intelligence organizations were using at that time. But in every review, they would tell us that if we could add one more feature, they would start using our stuff. This was the first time I was getting exposed to the politics. Product development does not result in publications, and in any case, they strongly discouraged us from publishing whatever we could potentially publish. After 6 years, we realized that they had no intentions of ever using our work, and we called it off, but the damage was done.

One day, we were sitting in the department and thinking how to improve PhD program. The outcome was that there aren't enough students applying for PhD not just in IITK but across IITs. So we need to tell students about the advantages of doing Phd. Just the previous year, IITB had arranged such talks in many colleges, and we felt that we should do something similar. With the department now very rich could easily afford to go around and give talks. I volunteered. I was given one semester off. Every week, I will choose three colleges where I could visit on Thursday, Friday and Saturday. The criteria was that I should be able to go from 1st to 2nd and 2nd to 3rd college either by a couple of hours of car journey, or by an overnight train. 12 weeks, 36 colleges. I thought what I should do for the whole day besides giving a lecture on why they should do PhD. I came up with a plan that included one technical talk, one career talk, meeting with students, meeting with faculty, meeting with leadership, etc. I used this opportunity to understand technical education. What are the issues our colleges are facing. What kind of curriculum they follow. What is the pedagogy. How do they hire faculty. And so on. I realized that while resources were a big issue in many colleges, autonomy was an even bigger problem, and in many cases, it was just the lack of exposure to best practices. We created a program under which any faculty member of any CSE department could spend a month at IIT Kanpur. We will take care of local hospitality and the person could sit in classes, sit in some of our meetings, meet any faculty and find out how, in general, we did what we did. Unfortunately, never took off, and we had less than just a few visitors from one college from South India. The colleges did not want their faculty to be away for even a day, and certainly not when they had to pay for it.

What has always been interesting in CSE department is that I could go around and make commitments, even those commitments which will have financial implications, and the department would always support those commitments. They supported me not just when we had good bit of money but when we had very little money. The trust between the colleagues was the hallmark of the department.

After this semester of traveling, I realized that very small things can improve the quality of education in India, at least CS education at college level. And I started writing articles in media, my website, and later started my blog. The reaction to most of what I wrote was that all this is good in theory but won't work in practice. That despite that semester on travel, I still haven't understood Indian higher education. So when I received the offer of leading LNMIIT, I grabbed it with both hands, and the institute became my lab. In 2 years, we did a lot of interesting things and looking at the success, I got even more convinced of my ideas.

One of the big part of department culture is its faculty tea room, a place we could go to any time and help ourselves with a cup of tea or coffee, all at the department cost. This was the place where we could discuss our ideas, discuss issues related to courses, students, and what not. It was a comfortable place to have discussions with a guest. But alas, one of the recent Heads stopped free tea and now we need to pay. This has ensured that only a few coffee addicts like me go there, and none of the department issues get discussed there (because what is the point of discussing with a small group, when a larger group would take a decision later on).

24 years is a long time. Lots of memories. I have not written about students at all. But it is already too long. May be if I am around next year, will write about students. Thanks for reading.