Data Science for Social Good — identifying students at risk of missing out on higher education

Datafest, a recent event at The Shard in London, marked the culmination of a twelve week effort by students participating in the 2023 Data Science for Social Good (DSSG) Summer Fellowship programme at the University of Warwick, supported by the London Mathematical Laboratory.

At Datafest, students reported on four specific projects which were undertaken this summer. In previous blog posts, we reported on several of these projects which focused on topics such as improving predictions of deforestation in Brazil, or helping consumers identify greenwashing in corporate advertising.

Here we’ll offer a brief summary of progress made on another project aiming to help students realise their full academic potential.

Why don’t more students pursue higher education?

University enrolment in the UK is now higher than ever. Even so, there’s reason to believe that the numbers could be significantly higher, and that many qualified students who could benefit from further education do not pursue it. According to Senna Ma, a project manager working on the problem, some 40% of students who meet the requirements to go on to higher education fail to do so.

“This was one of the big problems that United Learning, our partner wanted to look into,” says Ma. United Learning is a multi-academy trust formed from a group of state-funded and fee-paying private schools operating in England which oversees 93 different schools and 60,000 plus students.

“There’s a lot to keep track of,” says Ma, “and what they want to focus on is how to help students make the most out of their academic potential.”

In particular, United Learning wants to use data to learn the best forms of support so they can invest in or research groups they can create to help the students.

“It’s important to emphasize,” says Jack Buckingham, the technical mentor for the project, “that we’re not trying to say everyone should go for higher education. There are many valid alternative career paths.” 

The worry, as he explains, is that many students simply do not have sufficient awareness of their opportunities, which prevents them from making the best decisions. A student, for example, who may be the first in their family to go to university probably hasn’t had many discussions with their parents about their future possibilities.

“We’re trying to basically level the playing field,” says Buckingham, “rather than saying everyone should go for higher education.”

Mark Buchanan: Senna, how did you approach this problem using available data?

Senna Ma: “The initial step we took was just to better understand problem. And part of that is just knowing how much data we have access to. After discussions with strategic directors within United Learning, and professors studying and researching similar topics, we set out several objectives.”

A first goal, she says, was to predict the probability of students continuing their schooling and attending sixth form, which refers to the final two years of secondary education in England, Wales and Northern Ireland. A second was predicting the probability for students to obtain low, medium, or high scores on the General Certificate of Secondary Education (GCSE) exams, which are commonly taken in year 11 and play a primary role in determining students’ chances to continue their education.

The data scientists working on this project focused on making predictions for these students at year 9 and onward, as this moment, a few years before the GCSEs, appears to be particularly critical.

“Our study looked at data starting from year 9 because it’s a pivotal moment,” says Ma. “Although it’s very early, it’s a vital moment for a student to decide which subjects they’ll take. Some higher education programs require certain sibjects to be examined on their GCSEs.”

A third, critical goal, says Ma, was to identify students who would most benefit from additional careers conversations based on these predictors.

“We also wanted to identify common features, shared across these high priority pupils, so as to find ways to give them better support.”

Mark Buchanan: I understand your results so far are preliminary, but what have you found?

Senna Ma: “Some of the important features we found were things like whether or not a student is eligible for people premium grant,” a type of funding designed to improve educational outcomes for disadvantaged pupils in state-funded schools. “Or, a second important feature is if English is a student’s second language.”

As United Learning and other data scientists carry this work further in future, they hope it will show how data analysis can help educators spot the most crucial factors which prevent many students from being aware of the opportunities they have for higher education, and to design support programmes to help students who are interested in pursuing higher education.

Leave a Reply

Your email address will not be published. Required fields are marked *