Scientists use AI to develop better predictions about why children have difficulties at school

Scientists who used machine learning – a type of artificial intelligence – with data from hundreds of children with learning difficulties, identified clusters of learning difficulties that did not fit the pattern. previous diagnosis given to children.

Researchers from the Cognitive and Brain Sciences Unit of the Medical Research Council (MRC) at the University of Cambridge reinforce the need for children to receive a detailed assessment of their cognitive skills in order to identify the best kind of support.

The study, published in Development Science, recruited 550 children who were referred to a clinic – the Center for Learning and Memorisation of Attention – because of their difficulties at school.

Scientists say that most of the previous research on learning difficulties has focused on children who have already been diagnosed with a particular diagnosis, such as attention deficit hyperactivity disorder (ADHD), autism spectrum disorders or dyslexia. By including children with any difficulties regardless of diagnosis, this study has better identified the range of difficulties and overlaps between diagnostic categories.

Dr. Duncan Astle, of the Cognition and Brain Sciences Unit of the MRC of the University of Cambridge, who led the study, said: "Receiving a diagnosis is a landmark important for parents and children with learning difficulties.This diagnosis recognizes the difficulties of the child and helps parents and professionals who work with these children on a daily basis find that careful labels do not not counting their difficulties – for example, a child's ADHD is often not like another child's ADHD.

"Our study is the first of its kind to apply machine learning to a wide range of hundreds of distressed learners."

To do this, the team provided the computer algorithm with extensive cognitive test data from each child, including measures of listening skills, spatial reasoning, problem solving, vocabulary, and memory. . Based on these data, the algorithm suggests that children best fit four groups of difficulties.

These groups aligned closely with other data on children, such as parents' reports on their communication difficulties and educational data on reading and mathematics. But there was no correspondence with their previous diagnoses. To test whether these groups corresponded to biological differences, the groups were compared to brain MRIs performed on 184 children. Clusters reflected patterns of connectivity within parts of the child's brain, suggesting that machine learning identified differences that partly reflected the underlying biology.

Two of the four groups identified were: difficulties with working memory skills and difficulties in processing sounds in words.

Difficulties related to working memory – short-term retention and manipulation of information – have been linked to difficulties in mathematics and tasks such as the following list. The difficulties of processing sounds in words, called phonological skills, have been linked to reading difficulties.

Dr. Astle said, "Previous research on poorly readable children has shown a close connection between the difficulty of reading and the problems of treating sounds with words, but by examining children with a wide range of difficulties, we discovered unexpectedly The difficulties of processing sounds in words do not only pose problems of reading, they also have problems of mathematics.

"As researchers studying learning difficulties, we need to go beyond the diagnostic etiquette and hope that this study will contribute to the development of better interventions more specifically aimed at the cognitive difficulties of patients. children. "

Dr. Joni Holmes of the MRC's University of Cambridge MRC Brain Science and Cognition Research Unit said, "Our work suggests that children who find the same subjects difficult might struggle for very different reasons, which has important implications for the selection of appropriate interventions. "

The other two groups identified were: children with extensive cognitive difficulties in many areas and children with typical cognitive test scores for their age. The researchers found that the children in the group whose cognitive test results were typical of their age may still have encountered other difficulties affecting their schooling, such as behavioral problems, which did not occur. been included in machine learning.

Dr. Joanna Latimer, MRC Head of Neuroscience and Mental Health, said, "These exciting discoveries are at an early stage and are starting to look at how we can apply new technologies, such as machine learning. To better understand how the brain works, the MRC is funding research on the role of complex networks in the brain to help develop better ways to support children with learning disabilities. "

Explore more:
The hidden condition could be the real reason why many people have difficulties with mathematics

More information:
Duncan E. Astle et al, Remapping cognitive and neural profiles of children struggling in school, Development Science (2018). DOI: 10.1111 / desc.12747