Data in schools is heading for a revolution that could end up transforming how everyone in them works.
To understand just how big this change will be, consider our early attempts at using data in education as being like cave paintings, with students’ progress recorded with pen and ink in mark books.
Today’s league tables and government data can be considered more like photographs, offering a detailed static image.
But new technology allows data from schools and other education organisations to be turned into something comparable to the moving image, by facilitating the capture of a far higher volume of constantly updated information and providing the tools for instant analysis.
The implications of this new technology are enormous – it could see software packages take over a substantial element of the role currently carried out by deputy headteachers. It is also potentially controversial. One further education college is already using the technology to try to cut down drop-out rates by analysing the profiles – including the ethnic backgrounds – of students.
But the Association of School and College Leaders (ASCL) believes that there is huge positive potential for big data and is at the centre of a major project designed to ensure the new technology benefits schools.
Duncan Baldwin, ASCL’s deputy policy director, aims to combine hundreds or even thousands of schools’ management information systems to provide a more detailed, real-time picture of UK education, to learn new insights into what works and to put schools in control.
“My ambition here is to find a way for schools to collectively pool and group that data, so we have a big data set that tells us what’s going on in schools at the moment,” he says. “What they normally send off is things like exam results – which tell you about a particular point of time – and about students who’ve left the school, by which time it’s too late to do anything.”
Current data collection is a one-way street that leads to the Department for Education, he says. “It’s being used to judge the school. I think it’s time to switch the balance.”
By combining schools’ disparate sources of information and using high-tech methods of analysis, ASCL hopes that the data project could become a sort of national laboratory for education, answering questions on issues as diverse as the best order in which to teach topics or how to timetable your most experienced staff effectively.
It is an approach already being deployed by educational apps that have the ability to track a student’s usage and progress, such as Khan Academy’s learning library app or the Duolingo language learning app. As students increasingly work on computers and mobile devices, similar opportunities may open up for schools.
Duolingo, for instance, was able to use data to help remove an unexpected stumbling block in language learning.
It found that Spanish speakers learned English pronouns more effectively when they were only taught “he” and “she” at first, in gendered forms that they were familiar with, and only later introduced to the unfamiliar neutral “it”.
By combining disparate data sets, school leaders could gain instant new insights into how to operate efficiently.
They could, for example, calculate the cost per minute of teachers – or even the total cost of teaching individual students (see box, below).
Working with data experts, ASCL has been able to successfully combine data from about 95 per cent of the management information systems used by schools. “Getting the data out is solved,” says Mr Baldwin.
Now he wants to attract more schools to volunteer data and to help steer the project.
“It’s used very widely in more or less every sector these days: in retail, healthcare, sport.
“Where it’s not used thus far, sadly, is education. While there are plenty of examples of big sets of data, they’re not actually accessible to the profession,” he adds.
Mr Baldwin envisages the formation of a new charity to handle the data set, meaning that no commercial firm would have direct access to it.
Data-sharing protocols established for the national pupil database could provide a model for opening up the data to the right organisations.
“I can absolutely envisage this as a research environment for carefully approved organisations that may want to explore this as a tool for research,” Mr Baldwin says.
“The Commons Science and Technology Committee says the biggest risk with big data is that we don’t do it.”
According to Peter Burrows, chief executive of technology company DataSec Education and an adviser on the ASCL project, believes complex algorithms and artificial intelligence programmes could take on a deputy headteacher’s job of data analysis.
Systems such as Microsoft’s Azure (see box, above) and IBM’s Watson, the artificial intelligence (AI) system that “won” the US TV quiz show Jeopardy! in 2011, don’t just crunch numbers, they can also trawl through text, such as teachers’ remarks on a department’s progress, and combine this qualitative commentary with the hard data.
“If you look at individual schools right now, most have an assistant headteacher or deputy headteacher who’s spending 30 per cent of their time looking at and analysing data,” says Mr Burrows.
“That’s a huge amount of money – if you think that the average salary of a deputy headteacher is £50,000-£60,000 a year – £20,000 is being spent on data analysis by just one person, and that’s without the rest of the senior leadership, governors and middle leaders all doing their analysis at the same time.
“We’re talking about the probability of saving huge amounts of money and time by having developed AI systems that can do a lot of this,” says Mr Burrows.
“And the first step is to do it at a national level. Then we can take those algorithms that we’ve developed, those concepts and move them down the chain.”
Mr Burrows acknowledges that the stakes are high, with a need to protect data from misuse, ensure that the analysis is accurate and maintain a focus on educational improvement rather than accountability to external authorities. But he and ASCL believe that by ensuring that schools lead the project they can preserve its integrity. “People’s careers are on the line, if the analysis is flawed or incorrect in any way,” Mr Burrows says.
He believes that a high-tech data solution could be more reliable than humans, however. “Taking the human factor out of it and bringing in AI would benefit a lot of people, because it would be a fairer, more balanced method of assessment of teacher performance, pupil performance and, dare I say it, parental performance,” he says.
Viktor Mayer-Schönberger, a professor at the University of Oxford and co-author of the book Learning With Big Data: the Future of Education, says that schools need to protect the data and also convince parents that it is being protected. In 2013, a big data initiative that had been backed by the Gates Foundation in the US was put on hold in six of the nine states involved after a parental outcry about data security.
And Professor Mayer-Schönberger warns against data being used to select and reject students instead of improving education for all. “The stupidity of big data would be to use it as a souped-up or turbo-charged filtering tool, filtering certain students out and giving them even better education, and filtering laggards out, or filtering bad teachers out,” he says.
With these few caveats, Professor Mayer-Schönberger says that he would encourage schools to pursue big data projects.
“You can use these patterns in order to stimulate questions that you never thought of before,” he says.
“I’m very much in favour of it because I’m convinced it’s going to come anyway. Getting in early helps the stakeholders who are at the core of this, whether it’s students and parents or teachers and schools, to actually shape it, rather than having it imposed by society or by the government. It guarantees that this really becomes an empowerment tool rather than a filtering tool.”
A tool to tackle drop-out rates
One area where artificial intelligence and big data are coming together to help educational institutions make decisions is in college and university admissions.
Microsoft’s Azure platform has been used in further education institutions to crunch historic data on completion rates, enabling colleges to predict whether applicants will finish the course.
“We’ve been working with some colleges where they’re looking proactively at whether someone coming on to the course is going to be able to pass,” says Steve Beswick, senior director for education and charities at Microsoft. “It’s machine learning, where the computer looks at data and tries to predict the future. You can take data like ethnic background and you can look at income, the results that they got at A level, and when they come to college or to university where they’re paying their £9,000 a year, you can then predict what that drop-out rate will be.”
Software that can fit together a school’s data jigsaw to save time and money
In its work with multi-academy trusts (MATs), the financial software provider PS Financials has begun to explore the possibility of combining the large amounts of data that schools currently hold in separate systems.
Will Jordan, education sector manager, said that the company had carried out an informal poll of about 50 school and MAT leaders, which showed that they have an average of 14 different systems to store and analyse data, including assessment, asset management, budgets and strategic forecasting.
None of the trusts surveyed could calculate how costs and outcomes varied across their schools.
Less than a quarter of the schools and none of the trusts could calculate the difference in student outcomes based on teacher experience, although three-quarters said that they would like to.
Mr Jordan said that his company’s work with MATs had so far focused on combining timetabling information with financial data.
In one school, the research found that Year 7 teaching was costing 25 per cent more than Year 9 teaching because of the use of senior leaders as teachers and smaller class sizes.
Mr Jordan says such information can help schools by “making better use of the teachers”.
“Why are you wasting your time supervising an isolation room when that could be done by someone else who isn’t costing us that much?” he says. “It’s about understanding what wasting their time costs.”