Random Thoughts


5 Remote Micro-Internships Available

posted Jul 26, 2017, 10:41 PM by Prashant Bhattacharji   [ updated Jul 26, 2017, 10:42 PM ]

5 micro-internships for BTech/MTech/PhD students proficient in Python/R/numpy/scipy/sklearn-10k for about 60-70 hours of work.
The tasks involve
- Analyzing data and creating visualizations
- Writing basic Python code for simple tasks and then constructing tutorials around those. The topics revolve around ML, Data, Statistics
- Creating good tutorials on technical topics

You can work remotely, from your home or college. The work may be stretched over a period of 4-6 weeks, depending on your schedule and bandwidth.
People with github profiles or blogs will be preferred.

If interested, mail: updates@thelearningpoint.net

Literature degrees do not lead to unemployability. Over-Specialization does.

posted Jun 27, 2017, 11:12 PM by Prashant Bhattacharji   [ updated Jul 6, 2017, 12:20 AM by Prashant Bhattacharji ]

Sundeep Khanna write in Livemint about how the mismatch between Literature grads and 21st century workforce requirements.


Language adds color and decor which may be ostensibly pointless like the lights behind the fishbowl above. But language might very well help in communicating a point which might not have been conveyed effectively otherwise. But, let us see what Sundeep says. I agree with the points he makes, but the primary issue, is a mediocre and over-specialized curriculum and not the degree itself.

"The honours course in English doesn’t incorporate conversational English or even business writing but largely English literature as part of which students will learn to appreciate Jane Austen’s Pride and Prejudice, William Shakespeare’s Othello, Charles Dickens’s Hard Times, besides many Indian authors and poets such as G.M. Muktibodh and Nissim Ezekiel. Hell, there’s also Alice Walker and Jean Genet for a global feel.
Which is all very well for developing the finer sensibilities and even for some fascinating conversations on a rainy day, but severely restrictive in preparing students for a future career outside of academics."

Just having one common layer of general and applied courses in all majors (whether Literature or Engineering) will help everyone a lot. A few compulsory courses spanning the natural science, mathematics, statistics, languages, philosophy, psychology, economics, history and law. These are necessary for everyone. That, in fact, is the old "liberal arts" model, somewhat in line with the Columbia core program.
For all the engineers with BEs and BTech degrees, which we annually churn out, how many people can build a simple Class 10 level program or circuit. There's only so far you can go by flinging degrees on people who shouldn't have been allowed to exit high school anyway,

The real issue here, is not the major, but the curriculum. Whether DU or CBSE, they should detoxify their English syllabus by purging out the works of third rate Indian authors whose works are most likely included only because they get to grab coffee with the right people in their cozy Delhi circle. Amitav Ghosh, Nissim Ezekiel, Anita Desai, Vikram Seth - what's the point in including t in their works in the syllabus? Might as well study some math or programming or economics instead of that. People whose books couldn't make it in the markets and sell on their own merits are artificially propped up by forcing them into text books to humor them about their writing skills. At the school level, ICSE gets it right. They don't include Indian authors just for the sake of it.

People will pay like anything for good English classes in India. With a little bit of exposure to somewhat applied courses, if even a handful of these grads somehow became more enterprising and started NIIT/FIITJEE like corporate chains, within 5 years the standards in English would improve dramatically. What's the reason for the disproportionate popularity of English Literature though, I wonder - beyond the top colleges? It is almost similar to our "engineering degree" mill. I don't see an obvious reason as to why it is preferred as a course of study to major in, over subjects like Psychology, Philosophy, Economics, Legal studies, Finance etc.
That said, some Literature courses for all, can be a great way to improve writing skills. Never underestimate the potential of a letter well-written, as Jane Austen says in Persuasion!

Ashoka University is perhaps the best place in India to study these subjects. They have a broad curriculum, the true liberal arts curriculum with multidisciplinary academics as a key theme. The article also makes comparisons with Literature in US universities. That, in my humble opinion, is not a relevant comparison. In an English-speaking country where most college graduates are prolific writers and fluent in English, Literature grads in US simply don't hold a currency so valuable in the surroundings they live in, in comparison to their counterparts in India. That valuable currency, is a command over English.



Discourse driven by analyzing data and stats helped in the grade inflation crackdown

posted Jun 24, 2017, 12:53 AM by Prashant Bhattacharji

This score inflation and assessment quality discussion this summer, is the beauty of opening up basic data and stats. And this is why I say that the RTI is a terrible law. Most of that data should be open, online and anonymously accessible to everyone. Folk from TV news channels, reporters, data analysts etc. everyone can scrutinize that data and make organizations behave themselves automatically.

In fact I'd even say that RTI was architected to hide data and to send out a clear signal that people will be killed for asking the wrong questions. Almost 60 RTI activists killed, because you can't access data anonymously. Even with the demonetization tamasha - if note exchange data had been dashboarded and monitored right from day one, bank managers wouldn't have dared to launder at the scale they did, for fear of being caught.
Coming to this specific case, the babu running CBSE is having a very difficult summer. Getting slammed on a daily basis this year. Either dragged to the court or being lynched by the media or attacked by someone in the central or state government ;-) And now that the Times of India has decided to track them literally on a daily basis, may God bless them.
First for inflating, then for not inflating, sometimes for the MBBS medical entrance, sometimes for school exams, marking snafus, totalling errors and now for a panic stricken statement that they will hold exams in February from the next year, to get more time for proper correction. And the latest basket of tomatoes thrown at them is a contempt of court case for making a statement out of mere frustration. Actually most of what is now tumbling out is a consequence of the inflation crackdown. Student deserving 75, carelessly awarded 65 and bumped up to 80, didn't really need to complain in the past.

The 90+ band in the ISC examination has also seen a noticeable withdrawal, particularly in English. And remember - this is while marks have still been inflated.
Makes one wonder how sobering the real results are.


Meanwhile, let's see how this crackdown on score and pass-rate inflation is playing out. A nice visualization of crashing pass rates in Quint.








Bashing the Bihar Toppers is a fad. Let's look at CBSE Practical Scores

posted Jun 19, 2017, 11:23 PM by Prashant Bhattacharji   [ updated Jun 21, 2017, 4:29 AM ]

It has become a fad for folk to bash the Bihar toppers. Bihar gets so much attention only because, well, it is Bihar.

Here's the histogram of practical scores in Physics and Computer Science in CBSE. Nearly everyone gets either a full score of 30 or a nominal one or two marks short of it. And yet, if you try to hire, you can't find folk who can build a trivial circuit or program or even write a single paragraph without making serious grammatical mistakes. Then there's ICSE where the pass rate which hasn't dropped below 98% for 4-5 years.
ISC Pass rate has never dropped below 95 in the past five year.




There is a compounding cost to all this inaccurate assessment and score inflation. Have a few key stage assessments but get them right. The folk will go on to enroll in shady colleges, then probably do some name-sake post graduate degree too - when in reality, just a high-school education would have been sufficient to reach that standard, if the bar had been kept sufficiently high. It is easy to make fun of Bihar. But at least their toppers have scores in the high eighties at max, unlike CBSE/ISC where the centum is the limit.

Now, the CBSE is pretending to hold an enquiry into all these marking snafus.
As long as a steady stream of reporters and journalists keep discovering the magic of data uploaded our old blogs, these folk are not going to have an easy time specially now that people are freaking out all over the country. Such a massive examining body (also running critical entrance exams now) should have been run by a team of assessment professionals, academics and statisticians. Definitely not some UPSC officer suddenly shunted into an academic body when his prime competence is most likely dishing out orders to semi-educated folk in rural areas.


The 99% DU cut-offs are most likely a marketing gimmick

posted Jun 14, 2017, 11:48 PM by Prashant Bhattacharji   [ updated Jun 14, 2017, 11:52 PM ]

It is that time of the year again when Delhi University cut-offs will be in the news.
One thing I have always suspected is that DU's high cut-offs in the first few lists are nothing but a big advertising campaign much of the time. So even some unknown college gets its name in the paper by keeping a 99-100% cut off and going to town with that wild number. Nobody looks at the numbers in their third, fourth, fifth list.
When unheard-of colleges keep 100% cut-offs the marketing intent is as clear as can be.
And people will run a free advertising campaign for colleges which do that by making a headline out of it by shouting "99% cut-off" when the fact is, that even with all this score inflation there are barely 250 kids getting that score in CBSE or ISC and many among them do not apply to DU. If you trace them you will see that some will go abroad and many to an IIT or BITs and now a few will go to Ashoka. Some will go to places closer to their home: Kolkata or Mumbai University.
I can understand 95, 96. 97 percent cut-offs. And when students from Tamil Nadu or AP throng a specific college like SRCC, even 99% makes sense. But as of now that is the exception, not the norm. Though it might quickly change this year as word is out in South India that there is an incredible offer in Delhi University where one can simply walk in with inflated marks.

Much ado about the IIT JEE

posted Jun 13, 2017, 3:28 AM by Prashant Bhattacharji   [ updated Jun 13, 2017, 3:30 AM ]

The results of the JEE are out and one can see interviews of folk who spent the last 2-3 years slogging and cramming for this examination. It is encouraging to see kids pursue options other than IIT despite the ranks to do so.
At least 25% of the topics in the JEE exam are completely irrelevant for students of most programs - or can easily be taught in a 1 semester course in the streams which need them (from what I remember, mostly Chemistry). On the other hand by now the exam should have included some necessary general skills. It badly needs to cover critical topics like language, spatial skills, statistics, major current events and some algorithmic thinking topics like simple graph theory examples (railway network). The IIT JEE syllabus can do with a lot less esoteric topics.
The point is not to select a different bunch of people, though some of that will happen as well. There will always be pressure given a finite number of seats, the idea is to use academic energy "economically" to prevent burnt-out before on-boarding. If academics don't do the reforms then the likes of Kapil Sibal will.
The quality of Indian academics and educators is on full display whether it is the school exams or the entrance exams like the IIT JEE or the NEET. Which is why the next generation will be even worse.

Discussions around Rote, Rigor, Testing and Marks

posted Jun 13, 2017, 1:53 AM by Prashant Bhattacharji   [ updated Jun 16, 2017, 5:47 PM ]

Standardized testing is always a controversial topic and the concerns over rote and rigor always arise in any discussion related to it. 
There are two separate issues and we need to partition them appropriately. When people complain about specific tests and examinations, such as the workload required by the JEE, the real issue to zoom into is often the curriculum prescribed for it or the poor design of the test itself. 

The test and curriculum design themselves.
We can all agree on that - in the Indian context, a lot of top-down centralization and standardization of curriculum and systems is a disaster for such a large country. The NEET is just a recent example of this, not just for its legal and operational issues, but also academic ones. 
And the extent of the recall based component makes our tests far more of a burden than necessary, often with retrograde effects. 

The second issue is the relevance of test scores and academic assessments.

Fine grained reading into these scores is pointless. But regardless of what the most seasoned of school educators might recommend with or without authority, the professors right from Delhi University to Harvard to Oxford, do look at the results in exams and assessments of some sort. Similarly the hiring manager in a research or technology or finance firm will almost always screen out candidates who have no trace of academic accomplishment on their resume. They will not be enamoured by a teacher certifying that the student has potential to learn - they will screen using marks or grades. They are the gatekeepers of the economy and the workforce which I referred to. And they very much care about academic records, except for outliers who have proven themselves in other ways.

Great test scores or GPAs don't necessarily give people a free pass to anything, but poor ones lead to almost instantaneous disqualification. 
While we shouldn't obsess over either, the "marks don't matter" is only as true as the "money doesn't matter" line of thought: to varying extents, they both do. 
And while they tell far from the entire picture, to suggest that they don't matter at all, is to de-value the work of the schools or individuals or institutions which perform well. 
To deny that reality and to perpetuate that hoax, is turning into a bit of a fad nowadays, while ignoring the disastrous consequences of these poor test scores on the careers of folk in govt schools with poor results. As a stark example: Only 219 out of around 30k MBBS seats in TN, went to kids from govt. schools. Less than 1%. What was the bottle neck - their low marks (for whatever reason). 
It is likely that a similar analysis will see the same kids blocked out of a variety of career options for the same reason. The wash-back phenomenon of teaching to the test is a real issue - but preferable to not teaching at all (current situation in India).

We cannot conflate life stories with data. Plenty of individuals who excel in academics will fail in the real world and plenty who fare poorly in tests and academics will be fine in the long run.
But the probability that those who ace their tests will end up better off, is very much for real till the world discovers a better currency to calibrate competence. 
Singapore, Hong Kong, US, UK, Taiwan - all have various kinds of testing and assessments. 
Finland does fine without it, but it is a rich and small country which can afford a very high per-child spend. Also the Finnish model results in "most do fine" outcome but it doesn't produce that many high-achievers.

This is specially true in a developing country like India where any non-exam centric process, which is ostensibly liberal, could turn too expensive (and hence illiberal). 
US universities or closer home, Ashoka University - they are willing to consider a multitude of activities and accomplishments, while using test scores and GPAs only as a coarse filter.
However those activities need a lot of supporting infrastructure or exposure which an average kid in India will not be able to pay for. 








The IIMs seem to be doing very little of domestic or International interest

posted Jun 7, 2017, 5:07 AM by Prashant Bhattacharji   [ updated Jun 7, 2017, 5:08 AM ]

I came across this absolute stunner while trying to benchmark data on my site with that of other sites in the country. IIM websites have less traffic than mine - which, I hasten to add, isn't some sort of felicitation of the eyeballs I have managed to pull, but rather, a revelation of how little global or domestic interest there is in whatever worthless crap IIM faculty publish or how useless their websites are in terms of the content they democratize.


I was specially amazed by the IIM Ahmedabad graph and just as I suspected, there seems to be a bug in reaching many of their pages.
It is actually quite funny in some ways. The data uploaded on my website has been scrutinised and cited in papers by UK/HBS professors and they backlink to it, which is why Google possibly pushes it up in rankings. But not from any of these low-grade local "policy academics".
There is a very real danger with "policy wonks" to the entire country - there is a natural self-selection bias in India where lots of mediocre people with limited insight, or achievements, but good intentions, like opportunities to parade like peacocks with big shots and leaders, and will try to convert fanciful insights from the domain of academics or idealism, into *LAW*. Such people after failing to make a mark in anything worthwhile often find their "true calling" in "policy space".


Here's a look at how IIMB traffic stacks up verus mine. And another graph displaying the stark difference between the traffic on the website of IITB and one of the IIMs. That however, is not surprising.
The IIMs are not really an internationally known name while the IITs are indeed recognized as good undergrad schools for the most part.




Right from the NEET to the JEE Mains to the Board Examination: CBSE's IAS Bureaucrats are an Unmitigated Disaster

posted Jun 7, 2017, 4:52 AM by Prashant Bhattacharji   [ updated Jun 7, 2017, 4:56 AM ]

Nowadays the medical MBBS entrance (NEET) is also conducted by CBSE and after their high school exam, it is turn for that too to be stuck in the courts because they used different questions for papers in different languages! I think they conduct the screening test for IITs also nowadays and again they did something like using multiple sets of papers with different questions (and no kind of statistical calibration).

This is a body which knows nothing about assessments and is run by some third-rate IAS babus, the last two were suspected to be extremely corrupt. People spend years preparing for these exams so fairness should be non-negotiable. After all this recent fuss and court cases, it seems like their priority was to hide their data before more walls instead of fixing processes. Well we got hold of a copy of it in any case and look at one example from their class 12 exam, scores in English.
So now we're on a new level altogether: not only are exam marks distorted, but there is also a new variable in how your marks will be edited - which zone you take the exam from.

My old posts from 2013-2015 are here:

The 4 graphs are for zone 4 (Mumbai-Chennai), zone 9 (Delhi) and for zone 6,7 and the rest. Zone 4 is one kind of histogram, zone 9 is another kind of histogram, and the rest are more like one expects, though the extent of the negative skew reflects fairly liberal assessment. So if you can add 10-12 marks to your score by taking the exam from a class 12 school in Delhi instead of Gurgaon! What kind of an organization has been handed over charge of such critical assessments.


Chennai Zone Candidates (Competing with TN Board)? Here's how the English Core Histogram looks

A very generous inflation and a rather high modal mark of 91.


Delhi Zone Candidates (Competing for DU Seats)? Here's how the English Core Histogram looks

A very generous inflation and a very large cluster at 95.



Other Zone Candidates (Competing for DU Seats)? Here's how the English Core Histograms look (liberal marking, but no score tampering)




Now here's the thing.
(a) India is a country focused on exams and assessments.
(b) But we cannot conduct a few key stage assessments properly.
Basically (b) leads to (a).
When you can't check a few critical checkpoints properly (post-high school, post-undergrad -- SAT/GRE) then you require duplicity of exams and multiple layers of exams.

Just imagine. now we see interviews of CAT/UPSC toppers who in their 20s, even 30s, being felicitated for clearing yet another exam at an age when people are building companies, doing high end research, complex surgeries and what not. Frankly, if someone has already been among school toppers and done well in university exams/JEE/NEET/CAT etc. and all the other half dozen worthless exams which exist in the country, why do you require yet another round of assessment. Why not call at least such people directly for interviews.

The Unacceptable and Criminal Blunders by the CBSE right from NEET to JEE to the Board Examinations

posted Jun 7, 2017, 3:58 AM by Prashant Bhattacharji   [ updated Jun 18, 2017, 7:03 PM ]

Nowadays the medical MBBS entrance (NEET) is also conducted by CBSE and after their high school exam, it is turn for that too to be stuck in the courts because they used different questions for papers in different languages! I think they conduct the screening test for IITs also nowadays and again they did something like using multiple sets of papers with different questions (and no kind of statistical calibration).
This is a body which knows nothing about assessments and is run by some third-rate IAS babus, the last two were suspected to be extremely corrupt. People spend years preparing for these exams so fairness should be non-negotiable. After all this recent fuss and court cases, it seems like their priority was to hide their data before more walls instead of fixing processes. Well we got hold of a copy of it in any case and look at one example from their class 12 exam, scores in English.
The 4 graphs are for zone 4 (Mumbai-Chennai), zone 9 (Delhi) and for zone 6,7 and the rest. Zone 4 is one kind of histogram, zone 9 is another kind of histogram, and the rest are more like one expects, though the extent of the negative skew reflects fairly liberal assessment. So if you can add 10-12 marks to your score by taking the exam from a class 12 school in Delhi instead of Gurgaon! What kind of an organization has been handed over charge of such critical assessments.


Chennai Zone Candidates (Competing with TN Board)? Here's how the English Core Histogram looks

A very generous inflation and a rather high modal mark of 91.


Delhi Zone Candidates (Competing for DU Seats)? Here's how the English Core Histogram looks

A very generous inflation and a very large cluster at 95.

Other Zone Candidates (Competing for DU Seats)? Here's how the English Core Histograms look (liberal marking, but no score tampering)






Now here's the thing.
(a) India is a country focused on exams and assessments.
(b) But we cannot conduct a few key stage assessments properly.
Basically (b) leads to (a).
When you can't check a few critical checkpoints properly (post-high school, post-undergrad -- SAT/GRE) then you require duplicity of exams and multiple layers of exams.

Just imagine. now we see interviews of CAT/UPSC toppers who in their 20s, even 30s, being felicitated for clearing yet another exam at an age when people are building companies, doing high end research, complex surgeries and what not. Frankly, if someone has already been among school toppers and done well in university exams/JEE/NEET/CAT etc. and all the other half dozen worthless exams which exist in the country, why do you require yet another round of assessment. Why not call at least such people directly for interviews.

1-10 of 47