It was during a governors’ meeting in 2018 that Carrie Matthews, head of Willen Primary School in Milton Keynes, had an epiphany.

The moment of clarity came when a governor pointed out that the school was failing to use its attendance data properly by only looking at past trends and performance; it was not using it for future improvements.

“There was nothing we were doing with that historical data to help those children [in the future],” Matthews admits.

Fortunately for her, the governor who had spoken up was an expert in artificial intelligence (AI), Raymond Moodley. And Moodley had a plan to use AI to get smarter on school attendance.

Using AI to monitor school absence

Five years later, as attendance has come to the forefront of the school policy agenda, Moodley is not alone in what he’s doing: AI as a tool to keep children in classrooms is about to go mainstream - and not a moment too soon.

Post-Covid, school attendance issues have been raised as a key challenge by both headteachers and policymakers. Everyone from the children’s commissioner to unions has spoken out about it, and this culminated in the Commons Education Select Committee announcing an inquiry into pupil absence.

“The last few years have seen a worrying trend of children missing more school than ever before,” Robin Walker MP, chair of the committee and a former minister for school standards, wrote in Tes.

He promised that the inquiry would “draw on the experiences of headteachers, charities and other experts to look for solutions”, including whether “providing breakfast clubs, free meals, after-school or holiday activities and other measures can help reverse this damaging trend”.

‘Every school should have, in theory, a virtual data analyst on demand’

These are all important areas to focus on but the experience of those at Willen Primary School may convince Walker and his committee to take a much closer look at AI.

Moodley, now chair of governors at the school, was a member of the Institute of Artificial Intelligence at De Montfort University Leicester (DMU) and remains a visiting lecturer.

After that 2018 meeting, he gathered up several years’ worth of anonymised past attendance data from the school’s management information system (MIS) and set about interrogating it to not only see what had happened but how it could inform future strategies to stop it happening again.

“I built an AI and data analytics software that could identify patterns within the data,” he tells Tes. From this, a clear problem was revealed: “Monday morning sessions were clearly the most missed.”

More importantly, the analysis also showed that the pupils who missed the most Mondays were also those most likely to become persistently absent.

For Matthews, this was a eureka moment. In the past, individual teachers may have spotted a trend of a child missing Mondays on an ad-hoc basis, but this new analysis gave her a full overview of school-wide absences and the long-term impact it had.

The school’s response was to make Mondays as enticing as possible. “We had Magic Monday and Muffin Monday and a Monday once a month when you could wear your own clothes - all fun and low-cost ways to make Monday feel more enjoyable,” Matthews says.

Targeted communications

This was just one part of the plan, though.

Because the insights from the data gave the school a much clearer idea of the pupils who were on track to become persistently absent, it could target communications with their parents accordingly.

“We were always communicating with parents before but with the data we had, we were able to be much more specific and say, ‘Johnny’s had X many days off this term and, based on that trajectory, this is what it will look like in three terms’ time,’ and we could compare that to how previous pupils with that attendance record had performed by the end of Year 6,” explains Matthews.

All the data on past pupils was anonymous, Matthews reiterates, and she says the idea was not to scare families but to show them the real impact that absence had.

“It was about them realising that it was worth trying to break that cycle,” she adds.

Targeted communications with parents is something that the committee has already received insights on, too, with a report last month saying that schools should be given more support in this area.

So, did the Willen Primary plan work?

“There was a 35 per cent decrease in Monday absences, and the number of children who were classed as persistently absent went from 55 in 2018 to 24 in 2019,” says Matthews.

The wider impact of this was that the school’s attendance rate went up by 3 per cent, taking it from below the national average to above it.

The fact that data-backed attendance planning had an impact comes as no surprise to Lal Chadeesingh, principal adviser at the Behavioural Insights Team (BIT).

Research at BIT shows the power that can come from engaging parents with communications about absences that harness real data to explain why attendance is so important.

“We ran a project in Bristol where we were sending messages out to parents at the start of each new term, reflecting back on the data from the previous term, and letting parents know how many days of school their child had missed,” he explains.

• How artificial intelligence can be a force for good in schools
• AI could help teachers spot learning difficulties
• Could ChatGPT mark your students’ essays?

Rather than expressing a pupil’s absence rate in percentage terms, it was provided in day terms, so parents realised exactly how much schooling their child was missing.

“Schools tend to express attendance as a percentage. This can be confusing and may not clearly signal a problem when one exists, since in a school context ‘90 per cent attendance’ sounds positive but it actually reflects 15 days of school missed,” Chadeesingh wrote in a co-authored blog.

The data from the Bristol trial, although limited by the onset of the pandemic, shows a clear impact: “This approach boosted the proportion of students keeping good attendance records (95 per cent or above) by 4 percentage points (59.5 per cent to 63.3 per cent),” Chadeesingh wrote.

While this work did not incorporate AI, Chadeesingh can see the benefits that it brings.

“I could imagine a time in the future when the type of nudge or message that goes out to a parent about their child’s attendance is informed by all this data sitting in the background,” he says.

Coming to a platform near you

The difficulty in scaling up this AI analysis approach, though, is time and expertise: not many schools have an AI expert as a governor.

However, commercial companies are already looking at incorporating into existing products the same type of service Moodley built.

It’s something we could soon see in school management information system (MIS) products, for example.

“The team is working on next-gen MAT reporting, which will show the whole-school attendance and highlight data patterns that may indicate an increased risk of absence,” Arfan Ismail, product manager at Education Software Solutions, tells Tes.

“There are several well-known markers for this - things like changes in behaviour or attainment - and so we’re talking about developing a system that can help provide teachers with the insights they need to help them anticipate likely absences, based on historical trends.”

‘The challenge for the sector is AI will throw up a lot of stuff, and it won’t be true, it will just be correlation’

Ismail notes that providing this sort of functionality should help schools to keep pace with government demands, too, after the Department for Education said last year that schools should conduct regular analysis of attendance data to identify problems and intervene sooner rather than later.

“Schools are expected to…conduct thorough analysis of half-termly, termly and full-year data to identify patterns and trends,” the government report states. “This should include analysis of pupils and cohorts and identifying patterns in uses of certain codes, days of poor attendance and, where appropriate, subjects which have low lesson attendance.”

Other MIS providers are also looking at this space, with Bromcom CTO Huseyin Guryel indicating the business has “plans to build an AI framework embedded in MIS/SMS to automate certain processes” that would include spotting absence issues that could indicate a safeguarding risk.

Meanwhile, the CEO of Arbor, James Weatherill, tells Tes that he thinks “there is a real bright future for AI” in helping to tackle attendance issues.

Reducing the burden

Coming hand in hand with better school data will be more national benchmarking: the DfE has created new dashboards that collate school attendance data to “identify national, regional and local trends, and patterns in school attendance”.

It claims that accessing this will give schools greater insights into their own attendance data by being able to spot if they are tracking above or below similar settings either locally or nationwide. The DfE also claims definitively that being involved will “not add to your school’s workload”.

How realistic is that, though? Having the right data is only half the battle. Being able to ask the right questions about it and pull useful information requires a certain skill set.

The software providers say this will be part of their thinking.

“The goal is not only to capture and present every single dataset but to lessen the cognitive overload on teaching and administrative staff by providing simple, actionable intelligence,” says Ismail. 

Weatherill concurs that in the future it should be possible for AI to do the heavy lifting in terms of data analysis so school staff can focus on what to do with the data.

“Every school should have, in theory, a virtual data analyst on demand that they could type or speak to and ask questions that would help everyone become a lot better at using their data,” he explains.

This could mean asking for insights on past trends on a certain cohort or pupil type, or which days of the week are most commonly missed, for example.

Whether it’s bought in or self-built, AI is undoubtedly going to transform what schools can do when supporting good attendance, says James Browning, chief operating officer at Academies Enterprise Trust.

As well as the Willen Primary example of predictive trajectories, he says there could be improved use of AI to spot correlations between “progress, results and attendance on a per-pupil basis” with recommended actions based on what has worked for similar pupils in the past - from learning pathways to tackling truancy-related issues.

He also says AI could piece together insights from different datasets in an MIS to flag a potential issue far quicker than staff may spot - something Weatherill has a perfect real-life example of.

“There was a school that noticed a pupil was regularly having fights on a Wednesday. Then they realised they were often happening at lunchtime. Then they looked at attendance data and realised he was often late on a Wednesday,” he explains.

Having worked this out, a call to the pupil’s mother revealed the parents had recently divorced and on Wednesdays he was at his dad’s house and not having breakfast, so was hungry by lunchtime and getting into trouble.

“All the events they needed were being recorded in the MIS but the teachers had to go into each section and put it together before they got to the parents - but the process could have been massively shortcutted if the system could flag it up proactively,” says Weatherill.

Data privacy and ethical questions

Schools may argue they have been here before when it comes to edtech making big promises to revolutionise day-to-day operations. From implementation to staff training - and in this instance, most obviously around data protection - edtech can often be far more complex than expected in terms of ensuring success.

With AI, significant fears around data protection have already been raised in relation to other education products.

Commercial providers that we spoke to for this article offered assurances that proper safeguards would be in place.

‘AI is only a tool. There is no replacing the human here’

But Velislava Hillman, a visiting fellow at the London School of Economics and expert on data privacy related to edtech, has a more general fear about AI use. 

“A predictive algorithm may generate a false positive or a false negative [and] may lead to using other sensitive information in combination, such as disciplinary records, parental background, race, gender,” she tells Tes.

Hillman says this, in turn, opens up the risk that because AI will not understand context or have the “human” information a teacher holds that may explain absences, certain pupils would become subject to “heightened scrutiny and surveillance” that could then “risk marginalising and perpetuating injustice further” on these pupils.

Weatherill acknowledges this could definitely be an issue. “I think the challenge for the sector is AI will throw up a lot of stuff, and it won’t be true, it will just be correlation,” he says.

He adds that MIS providers must work to build schools’ trust in these tools, such as by ensuring transparency from AI about the data it has used for any recommendation about a pupil.

Secondly, he says that because “AI is not sophisticated or tested enough to make decisions in schools”, it must be made clear the tool is only there to provide insights and recommendations. “The actual decision on what actions to take should be made by people with real-world experience,” he adds.

The power of the teacher

Hillman agrees that teachers must remain central to any decisions to ensure that any use of AI is ethical and proportionate. “Human involvement must be at the start and end of an algorithmic output,” she says.

Matthews’ own experiences underline exactly this: while it was the AI that spotted the trends, it was the human element - understanding how to make Mondays more enticing for pupils and deciding what to say in tailored communications with parents - that made the difference.

“AI is only a tool and the human professional will always have the final decision-making authority on the implementation and actions to be taken as a result of recommendations by a tool,” she says. “There is no replacing the human here.”

Yet while it is not about replacing the human, she says there should also be an openness to the power that AI can bring to help tackle the vital issue of improving school attendance.

“AI absolutely transformed the way that we looked at attendance data,” she says.

“That mindset change meant we were looking at attendance before it became a problem - rather than after it had become a problem.”

Dan Worth is senior editor at Tes