AI, it seems, is still a somewhat foreign and perhaps even intimidating concept. An engineer was recently fired for claiming an AI system was sentient, sparking fierce (unfounded) sci-fi doomsday predictions. In reality, AI is not something from a Marvel film; it is a tool to increase efficiency and increase results – helping humans to analyse data.
Usage is increasing fast, and a recent McKinsey report found that AI could potentially deliver additional economic output of around $13tn by 2030. Within the majority of industries, the use of data and machine learning is commonplace, and essential. Why, then, is education lagging behind?
The approach to AI in education
Policy formation in education tends to be based on meta-analysis (merging findings from various studies to produce a result) and generalised data. The issue with this is that data set combination is blind to one basic fact – context matters. It’s an obvious point to make, and yet one that must be made: what works at one level is not guaranteed to work at another.
Certain academics’ work, notably John Hattie, has frequently been cited as justification for this approach to policy at all levels, systemic right through to individual school class; others, however, have pushed back on this. Robert Marzano describes in School Leadership that Works how their initial findings from the meta-analysis saw strength of correlation vary across studies. Comparing non-American studies with American studies skewed the results – as they didn’t account for the level of training and qualifications that American school principals had. In essence, data is at its best when it is localised.
Using data effectively in education
Using data effectively in education is something we must do – and management information systems providers have the potential to be in the vanguard of those addressing questions around data use. Initial research from Education Software Solutions (ESS) suggests that while schools excel at collecting data, they are perhaps missing opportunities to put this data to use. The UK education system should be infused with data, and there are ways this can come to be.
With data, the timeless adage ‘quality over quantity’ holds true – having more data is no measure of success or results. Often, a deep data pool can be counter-productive, as with more information to look at, the attention is drawn to areas which have no direct relevance to the task. Statistical relationships between intervention and outcomes are not always important. Education, as readers will know, is not infinitely resourced – indeed, quite the opposite. The game-plan, then, must be to refocus resource to ensure that actions being taken are impacting student outcomes.
“The human mind (powerful as it is) struggles to understand large-scale, data-heavy questions of causation”
One piece of research (carried out by Fazila Duyan and Rengin Ünver) found students focused more on a purple classroom wall than on a red one. What do we take away from this? That it is time for a refurbishment? That the overalls and paintcans should be brought forth with all haste to recapture the attention of the youth? Hardly.
Interventions that are implemented to improve the system have to have a high return on investment (RoI), and in order to make this the case, data analytics and handling must be improved.
Is there a solution?
How can we transform the way schools use data?
The answer lies with AI and, in particular, machine learning. One way to successfully use AI is posited in Robot-proof by Joseph Aoun, president of Northeastern University.
He suggests marrying together the processing and analytic capability of machines with the creativity of humans and our capacity to identify problems. This way, the problem is identified by human intellect and solved by machines.
The human mind (powerful as it is) struggles to understand large-scale, data-heavy questions of causation. This is unsurprising – AI, and in particular machine learning, should be used to trawl through the data and calculate which factors are critical and how spend corresponds to RoI, allowing for more advanced solutions.
In education, children’s outcomes (and safety) are paramount. Data products that prioritise these two things can help solve numerous problems within education. Predictive and prescriptive analytics, for example, can help target attendance-related issues such as persistent absence. Data-engineering and data-science, combined with in-depth, first-hand knowledge and experience of education within the UK on a day-to-day basis, is an essential combination in the push to build data-reporting software.
Intelligent insights will help to better student outcomes, with AI transforming general and abstract transactions into something idiosyncratic – bettering education establishments, from MATs to schools.
The most exciting and ultimately life-changing possibility inherent in AI use within education is specificity of learning. A future where every child can receive personal support and targeted interventions, helping to keep them safe, fulfil their potential and achieve their best throughout their time at school, is within our grasp – machine learning is the tool that will make this future a reality.
You might also like: Blackheath deputy head: ‘I can’t see any downside to having diversity of thought’