My summer internship at CSSL
By Shubham Kumar
As we might imagine, the life of an intern is not just photocopying and coffee
runs, or that one-off assignment you got because no one else wanted to do it.
Infact, it was quite the contrary at my summer internship. To use the cliché –
it was action packed!
Call me a geek, but I have always been mathematically inclined, I
find numbers and data exciting. But the dilemma, like many
others of my generation have, is — how to apply these skills to
solve real world problems.
In part, it was the pursuit of an answer to that, which brought me to the
Centre for Science of Student Learning (CSSL). Using diagnostic assessments
to improve student learning outcomes sounded like an interesting concept. So
part curiosity, part opportunity, and I was on my way to Delhi from Kolkata.
Now first things first, if you ever tell anyone you are going to Delhi in May,
they start believing you have self-destructive tendencies. Summer in Delhi is
brutal. It is not a good idea, not by any stretch of the imagination. But frankly
it didn’t matter. The offer from CSSL was to work with the core Research &
Development team and even the searing heat wasn’t going to keep me from
that.
On my first day, I landed at the CSSL office nervous and a bit unsure. No one
asked me to make coffee. They liked to make their own. It was an excellent
start.
I was instead initiated into the R&D team and it was serious business from the
get go. I would be working on psychometric research. Simply put –
psychometrics is the measurement of mental abilities, traits, and processes.
CSSL has been working to forge new paths in psychometric research by
leveraging psychology, statistics and technology.
The action began quite early. Within my first week, I was working on data
collected from assessments conducted in Rajasthan. My first job was a
challenging exercise — to develop a fuzzy name matching algorithm to find
common students who appeared in three different tests in the state, so that we could do a combined analysis of their performance.
Now here’s the thing — your first tryst with real data can be both exhilarating
and terrifying. It’s like having a key to many stories, questions, and answers,
and some of those will be novel. You might discover things that no one else
has. But it also means it has to be managed with caution. These are not just
numbers, but there are people associated with those numbers.
As part of the task, I developed an algorithm and created an R package with
the help of my mentor. And in the process picked up new concepts in coding
and data mining.
Of assessments
The most insightful thing I learnt while working at CSSL was the significance of well developed assessments. I realized that the worth of any educational assessment endeavour depends on the instruments i.e. the tools and techniques used. If these instruments are poorly designed, the assessment can be a waste of time and money. Secondly, how much students score, or what their grade is, does not help us understand how much the student actually knows, or how much they have learnt.
Sample this — Two students score 9/10 in a test. One of them answers a very
easy question incorrectly, while the other could not answer a difficult question correctly. In that case, do we assume that both students performed equally well? Many tests also give different questions to different students, are their scores comparable?
All these questions led me to a point where I was beginning to see how
research and data helped find those answers. I learnt about Item response
theory (IRT), for example, which we were using to diagnose gaps in learning.
IRT involves diagnosing the response to individual items or questions in a
test, compared to focussing on the test as a whole.
Apart from all the other things I got to do, working on a cheating detection
algorithm along with my team to detect copying in large scale tests was by far
the most interesting. We were using it to detect suspiciously similar response
patterns between students, who took multiple choice tests. Using advanced
statistical concepts and computational coding we developed sophisticated
packages that could basically figure out if students cheated.
The two months I spent at CSSL introduced me to a whole new
science, the Science of Learning where I saw people from multiple
backgrounds work together using diagnostic assessments as a
starting point and derive meaningful patterns from data to solve
the learning problem.
My time here was fascinating and helped shape my understanding of data
science involved in the education sector, something I would like to explore as
a career.
Shubham is a first year student of M. Stat at the Indian Statistical Institute,
Kolkata. He did a two-month internship with CSSL, and spent his time
working with the core Research & Development team.
Excited about the role assessments & data science can play in revolutionising how children learn? We’re hiring Interns, Resident Experts and Researchers in different subject areas. Apply to our open positions here.
