It’s not so long ago that if we heard the term artificial intelligence (or AI) it would conjure up thoughts of science fiction movies and a future that seemed a long way away.
Humanoid robots may not yet be a feature of everyday family life, but advances in machine learning, a form of artificial intelligence, mean that it’s already well advanced in a number of different sectors – from customer chatbots to combating cyber warfare.
What is machine learning?
Put simply, machine learning is when computers are able to learn from data without being specifically programmed, including being able to make predictions from that information. This type of business intelligence is also known as predictive analytics and is now embedded in our everyday lives, providing services we take for granted. If you’ve ever been recommended music or a book based on your buying trends, that’s predictive analytics at work.
However, as Neil Mackin, analytics director at Capita, explains in his earlier article, machine learning is about more than the personalisation of products and services – he says that the real added value is when it’s embedded into everyday processes and activities, making them ‘better for humans’.
It’s not, of course, just commercial organisations who can benefit from good business intelligence. Using machine learning, local authority teams can apply knowledge of the early signs of vulnerability to identify children who are at risk of negative outcomes.
How machine learning can help teams to protect vulnerable children and young people
Let’s say Sam, Holly and Alex, three fictional 16 year-olds living in the same local authority area, were each recorded as being Not in Education, Employment or Training (NEET) this year. Closer examination of their individual situations might show that Sam and Holly hadn’t been attending school regularly, and that all three young people showed much lower levels of progress than expected. Their records might also show that Alex had special educational needs and was finding it hard to stay engaged with learning. Added to this, Holly and Alex had both changed secondary schools several times.
With machine learning, a computer can look at these factors and large numbers of others which reflect similar characteristics, taking into account all the key events over the course of a child’s life and not just current indicators. It can then indicate which of these repeatedly come up for young people who eventually become NEET and what significance or weighting they each have on the outcome being investigated.
Where humans might consider the characteristics of a child at that particular point in time, computers can cross-reference anything and everything that’s happened in that child’s life, looking for relationships between the data, trends and probability. The ability to look at all these characteristics at once – in some instances this could be, say, 16 years’ worth of data on a young person – can highlight indicators that may have happened early in that child’s life, perhaps that they hadn’t attended pre-school in their early years, or displayed poor attainment at primary school. This is critical information in enabling local authority teams to understand where they can proactively target services much earlier and to greater effect for improving outcomes for that child when they reach their teens.
In summary, the advantage of machine learning is that it can consider a much larger number of variables than a human brain ever could, and without the unconscious bias that people may bring to the analysis based on their previous expectations of what might be significant.
In this way, models for different scenarios can be built using information about results we already know, to try and identify young people who are most at risk of falling into the same situation as Sam, Holly and Alex.
This can also be applied to youth offending and involvement in gangs, with local authority teams able to put in place early help measures to prevent young people ‘tipping over’ into committing an offence or joining a gang.
The future of machine learning in local authorities
By using machine learning to identify and monitor vulnerable groups, local authority teams can have access to high quality insights and predictions about what’s really happening in their communities, enabling them to make more informed decisions about early interventions.
Machine learning also has an important future in terms of improving the effectiveness of, and informing the commissioning of, services, by analysing factors amongst young people experiencing positive outcomes, to see what it was that ‘made the difference’ for them, and then make use of this knowledge so that local authorities can maximise those elements for other young people. Was it having the opportunity to play a lead role in a local youth drama group, be involved in a school band or receive counselling from a CAMHS therapist? Machine learning would be able to pull out the common factors so that they could be applied in other situations to successful effect.