Exam U-turn: behind the algorithm that triggered A-level grades mayhem
‘Seeds of the policy disaster’ sown in letter sent by Gavin Williamson as lockdown was announced in March
As anger erupted following the release of A-level grades last week, Prime Minister Boris Johnson insisted that the results were “robust”, “good” and “dependable for employers”.
But just days later, following Ofqual’s sudden withdrawal of the criteria for appealing grades, the government was forced into an embarrassing U-turn, with results now to be based on teachers’ predictions rather than those of a controversial algorithm.
The automated system was used in a bid to avoid what Education Secretary Gavin Williamson described as “rampant grade inflation” amid the backdrop of the coronavirus pandemic and cancelled exams. But who came up with the algorithm, which could end up costing Williamson his cabinet job?
The “seeds of the policy disaster were sown on the day the lockdown came into force”, when Williamson warned in a letter to Ofqual that avoiding grade inflation was a priority, The Times reports.
“Ofqual should ensure, as far as is possible, that qualification standards are maintained and the distribution of grades follows a similar profile to that in previous years,” Williamson told the exams regulator.
In other words, says the newspaper, “despite the fact that pupils would not sit exams, the government wanted to treat the class of 2020 like that of previous years”. A-levels were treated as the “‘gold standard’ of the education system and were not to be devalued”.
But in “commissioning the exams regulator to take out an insurance policy in the form of its ill-fated algorithm”, this “desire to cap grade inflation went too far”, adds the BBC.
What went wrong?
At the education secretary’s request, the regulator’s statisticians set about devising a system for handing out grades that “didn’t allow exam results to go up from previous years”, explains Jo-Anne Baird, professor of educational assessment at Oxford University and a member of Ofqual’s advisory committee.
The problem was that “in the case of the class of Covid, the preoccupation with maintaining standards came at too high a price”, according to the BBC.
A total of 39% of the A-level results released last Thursday were downgraded, and “pupils in disadvantaged areas were disproportionately hit the hardest”, says NS Tech, a division of the New Statesman.
The algorithm predicted grades after being fed various bits of data.
“The first was the teacher’s predicted grade for each student based on their performance in class and the mock exams,” the news site explains. “But this was deemed insufficient on its own, so teachers were also asked to rank each student from highest to lowest in terms of their expected grade.”
“Schools threw themselves into the assessment task,” adds The Times, with department heads leading meetings where “teachers argued the case for their pupils”.
But, says the paper, “there was a catch” in the Ofqual system. A report released by the regulator last week revealed that teacher-assigned grades were only given priority in classes of less than 15 students - a system that favoured private schools with smaller class sizes.
By contrast, for pupils in larger schools, “grades were far more influenced by the school’s historical performance and their teacher’s ranking than their predicted grades”, NS Tech adds.
This discrepancy accounts for the disproportionate number of students from schools that do not usually send pupils to the UK’s top universities who saw their predicted grades aggressively downgraded.
Was a fairer system possible?
According to The Times’ science editor Tom Whipple, “making a fair algorithm is like trying to unboil an egg” - that is, “impossible”.
The problem, Whipple writes, is that “when people extrapolate from population data to make predictions about individuals... you can end up making all sorts of counterintuitive, surprising and sometimes absurd mistakes”.
This is what went wrong with an algorithm based so heavily on a school’s historical results, he argues. “Clearly, this will be unfair to exceptional children in unexceptional schools”, while “conversely, it will be overly kind to unexceptional ones in exceptional schools”.
Sam Freedman, CEO of non-governmental organisation Education Partnerships Group, agrees with this verdict. The algorithm was “inevitably going to hit outlier students who were at the top of the distribution in schools that haven’t had many high performers in the past”, he tweets.
But, Freedman adds, the government’s decision to only use teacher’s predicted grades is also “unfair on pupils at schools who graded cautiously, unfair on past/future cohorts, and create[d] a lottery for uni places”.
And the U-turn may come too late for some students, with many universities saying courses for the next academic year are already full.
As for the algorithm, “statistics are by definition a way of representing many numbers in fewer numbers”, Whipple says.
“This is tremendously useful, but we need to know what it means: forgetting about the individual.”