Algorithms were once the little-known preserve of computer geeks, but today they control almost every part of our lives, from how we trade stocks and shares to how we socialise.
Now they are being put to use in a US school district, under revolutionary plans to identify children at the age of only 6 or 7 who are likely to drop out of high school before they reach 18.
The Montgomery County Public Schools district on the outskirts of Washington DC is developing an early-warning system that can pinpoint which of its first-grade students are unlikely to complete their school careers.
The algorithm was created by Thomas C West, a PhD student at the University of Delaware. Mr West said the system was made possible because the school board keeps detailed data on each of its students over the 12 years of their schooling.
He looked at the county’s senior high school class of 2011, from which 833 of 11,234 students dropped out. He studied their first-grade attendance records and school report cards to find common characteristics, and found that at least 76 per cent of the students who had left early could have been correctly identified as dropouts.
“We first looked at attendance, because for the most part first-grade students don’t miss school. They aren’t likely to play truant, so we looked at those children that had missed three days or more in one quarter,” Mr West said. “If they did then it tells us something about what is going on in their home life.
“We then looked at their behaviour and looked to see if they had been suspended. Again, it is very unlikely for a first grader to be suspended unless they have done something serious like bring a weapon into school. Finally, we looked at their reading and mathematics tests scores to understand their academic performance.”
What the number crunching showed was that a first grader was twice as likely to eventually drop out if they missed three or more days, or if they underperformed in reading and mathematics, but five times as likely to drop out if they had a bad behaviour record. The school district believes that this information will help it to target support at students from a very early age.
The UK government is also turning to assessment of children at the beginning of their school careers. The Department for Education is looking at introducing standardised tests for five-year-olds to enable more detailed evaluation of a child’s progress through school.
But although Mr West’s algorithm correctly identified more than three-quarters of dropouts, it also over-identified 47 per cent of students, as the risk factors for some would trigger red flags but then return to normal levels.
“What the initiative does is start the conversation between teachers, parents and students,” Mr West said. “Identifying the students is one thing; what is more important is putting in place the right interventions once we have identified them to keep them from dropping out.”
Russell Hobby, general secretary of the NAHT headteachers’ union in the UK, said every school should have indicators to show if a child is getting into trouble, but questioned whether the measure would improve on a teacher’s own judgement.
“Numbers can be useful at a very high level, but a teacher can generally spot if a child is having trouble at home or their parents are going through divorce,” Mr Hobby said. “Whether you need to put that into a cast-iron algorithm is a bigger question.
“The worry then is that it becomes a self-fulfilling prophecy, that if there are tell-tale signs then you automatically think a child is likely to struggle, or if there aren’t any then they won’t, which isn’t always the case.”
