Education Bubble in US - a response

A couple of days ago I had commented in this blog on Peter Thiel's article about education bubble in US. My friend Ajay Bharadwaj sends me link to a blog by Robin Cangie, where she argues that Peter Thiel is wrong, the situation is worse than a bubble.

Here are some of the points she makes:

"Students pay top dollar, not for quality, but for a name brand education. Classes, even at elite universities, can top 500 students and are disproportionately taught by poorly paid adjuncts and graduate students, not professors. Cheating and grade inflation are rampant and quietly tolerated. All of this points to a spectacular betrayal of the educational principles that these institutions are supposed to uphold – namely, to educate."

Was she writing about US education, I wonder.

At IIT Kanpur, undergraduate students don't pay top dollars or rupees, but they come not for quality, but for name brand education. More than half the students have no interest in the specific program they are pursuing, and will not get into a career related to that program.

Classes at IIT Kanpur exceed 500. For decades, we have followed the model of a large lecture followed by smaller tutorials. But it was a success when the class size was 200-300. Faculty does not necessarily believe that the model is scalable to 500+ size classes, but quality of education is less important than the faculty workload, and splitting the class into two will require one additional faculty member. (This was the reason given for large classes in a recent meeting that I attended.) The tutorial section size has gone up from 30 to 40. The tutors are increasingly graduate students, and not faculty, as they used to be. But we do not have any formal mechanism for training these students on how to handle a group of 40 students. (By the way, I am fully in favor of using graduate students for tutorials and labs, but someone should be telling them what to do.)

Cheating is not rampant, but is quietly tolerated. In last few annual reports of the disciplinary committee, there is hardly any mention of cheating cases. And now, it has been decided that in most cheating cases, the matter can be resolved by the instructor hirself, and it may not even be reported to the disciplinary committee.

Grade inflation has been encouraged in the last few years. The failure percentage now is little over 2 percent - mostly those who did not give the end-semester exam, or were ill. And the graduation requirement has been reduced from a CPI of 5.0 (on a 10 point scale) to 4.0 (from 2011 batch onwards). If you can pass JEE, you deserve to get a degree from IIT Kanpur.

And yet, I have no doubts that IIT Kanpur continues to be one of the best technical education providers in the country. I shudder to think what will be happening in thousands of other colleges across the country.

What is wrong with our Higher Technical Education

Everybody agrees that the higher technical education in our country is of very poor quality. The top 10 institutes provide very good quality education. The top 50 institutes provide good education, and everyone beyond that is bad. Talking specifically of education in IT related fields, different people talk about anything up to 80% graduates being unemployable (that is, they can't even be trained).

People from industry would love to argue that this situation has come about because academia does not listen to them. They do not have curriculum in line with industry needs. They don't teach them the latest technologies. They don't expose them to live projects. They don't put enough focus on non-technical skills like communication skills, teamwork and what not.

I have a very different take on this. It is not that these institutes (beyond top 50) are not teaching their students English, or Dot-NET, or any other latest technology. The problem is actually worse. They are not teaching them anything.

Recently, I had a chance to look at some statistics from the CS paper of GATE 2010. The average marks were 12 out of 100. This is after considering the lowest marks to be 0. (In reality, the lowest marks were -21.33.) So, if we consider real marks achieved, the average would be around 9. And the median would be even lower,
around 7 marks. More than 10% of the students had negative marks overall.

It set me thinking. What would happen if we were to ask 1 lakh 12th class students to give GATE. I would guess that unlike BTech 4th year students, the 12th class students would leave most of the answers blank (and hence avoid negative marks), and only answer those 5-7 odd questions which s/he is confident about. (There were a couple of easy questions on aptitude. There were a couple of questions on programming that 12th class students have been exposed to. And there were a couple of questions which had so much information given that one did not need to known any computer science to get those right.) My gut feeling is that if we were to give the same GATE paper to one lakh 12th class students, the median marks may be only a couple of marks less than what has been the case with these 7th semester students.

It means that 7 semesters of technical education has enabled our graduates to get 2-3 extra marks (out of 100) compared to what 12th class students can get. Remember, GATE paper is about basic computer science only. It is not about the latest technologies. It is not about industry trends.

So the problem is very simple. There is no education going on in thousands of colleges around the country. These students are not being taught even basic programming, or data structures, simple algorithms, basic computer organization concepts, etc. And hence any attempt to improve the employability of these graduates by training them in communication skills, dot-net, java, software engineering, and so on, is futile. One has to first see how we can ensure that they learn computer science basics. Unfortunately, I have no solution to offer.

Another interesting statistics from GATE 2010 paper was that in several questions, the average marks received by the students was negative. As people who have given GATE would know, you get 1 mark for the question if you answer it correctly, and -0.33 if you answer it wrongly. The scheme has been designed so that, if people were to guess randomly, then the average marks obtained would be zero. (Assuming, 1/4th of the students will answer each of the four options.) If some people know the answer and mark it correctly, some have left it blank, and others have given a completely random answer, then the average score should be positive. If we take out those who have not attempted, and those who genuinely knew the answer, and consider the rest, 75% of those should (statistically speaking) give a wrong answer. And if you add those who genuinely knew the answer, the percentage of wrong answers should be less than 75%.

But, in some questions, more than 85% people (out of those who have attempted the answer) have done it wrongly. This is too high a number (compared to 75%, in a sample of 1 lakh) to be considered a statistical anomaly. I discussed those specific questions with a few colleagues, and it occurred to us that the only reason why this can happen is if the students are not answering it randomly, but are confident of the wrong answer being right. Which means that they have been taught the subject matter of that question, but have been given wrong concept or information (which is worse than not teaching at all).

Recently, there was a proposal to have an exit test for MBBS to ensure that the degrees given by all universities in the country are adhering to a minimum quality for MBBS. I think there is a need to have an exit test for all BTech in this country. That will be a sure shot way to weed out poor quality institutions.