World

Britain’s Exams Chaos: The Fiendish Algorithm Behind Shock A-Level Results

Three days after Britons studying for their A levels learned their results, thousands of students are still trying to get their heads around the process that gave them lower-than-expected grades and dashed their hopes of getting into their dream university.
Sputnik

Britain’s latest national scandal isn’t about politics, as one might’ve expected.

Almost 39 percent of England’s A-level results – the qualifications needed to get into higher education – have been pushed down by one grade or more by an algorithm developed by the exam's regulator.

This has provoked widespread outcry from disgruntled students and their parents, as well as teachers and politicians.

The scandal will likely grow bigger as the regulator, Ofqual, is expected to award nearly 5 million GCSEs (lower-qualification grades) using the same new system.

 

Were there any exams in Britain this year?

The coronavirus pandemic, and the resulting lockdown, prompted British schools and colleges to cancel GCSE and A-level exams.

Instead, students received formal grades that fairly reflect their performance without actually sitting an exam – at least, that was the idea.

 

What was the way out?

Ofqual has tested 12 different statistical models and chosen one it called direct centre-level performance, or DCP.

It has given A-level grades to pupils using an algorithm that took into account teachers’ predictions, mock exam results, and the results of people who took the exam at the same schools in the last three years.

Apart from submitting predictions about how their students would have done if they had taken the exam, teachers also drew up rank orders of students for each subject from best to worst.

Ofqual explained that relying solely on teachers’ recommendations would have yielded “implausibly high” results – better than in any year on record – because schools and colleges tend to be optimistic about their students’ future attainment. Meanwhile, the regulator added, teachers’ judgements tend to be more accurate when they are ranking students rather than simply predicting their grades.

Britain’s Exams Chaos: The Fiendish Algorithm Behind Shock A-Level Results

“That seems to be a big inconsistency – you treat kind of one set of data, the rankings from teachers, as fixed and accurate, whereas the predicted grades are considered subject to uncertainty,” said Guy Nason, chair in statistics at Imperial College.

As the released A-level results showed, teachers in England had two in five of their A-level assessments adjusted downward by the statistical model. This amounts to around 280,000 students; and some of them are slated to miss out on their chosen university courses after not meeting the required grades.

 

What went wrong?

The problem appears to be with the algorithm moderating the results based on the A-level grades from the same schools from 2016 to 2019.

The idea was that a school is expected to perform the same on average in a subject as it has across previous years. This approach was meant to stop this year’s students from getting higher grades than the previous cohorts and retain the credibility of those institutions among employees and universities.

But it also appears to have punished students from lower-ranking institutions who did better than their school’s average.

Ofqual’s data revealed that private schools have fared much better, with their share of A* and As rising by 4.7 percent this year. This compares with a rise of 2 percent at state comprehensive schools and 0.3 percent at state sixth form colleges.

As many as 49 percent of grades at English private schools were A level or above. Conversely, such results were awarded to 21.8 percent of students at comprehensive schools and 20.8 percent at sixth form colleges.

The picture is also indicative in Scotland, where students from the most deprived backgrounds had their pass rate fall by 15.2 percent compared with a 7-percent fall for the wealthiest backgrounds.

“In summary, people who come from areas where people have scored low are assumed to score low this year, and people who come from areas where people have scored high are assumed to score high this year,” Ashwin Iyengar, a PhD student in pure mathematics at the London School of Geometry and Number Theory, told Sky News.

 

What next?

Scotland, where the exam body lowered nearly 125,000 predicted grades (26.2 percent), has already scrapped downgrades and replaced them with the higher estimates submitted by teachers.

In Northern Ireland, 37 percent of results have been downgraded. The education minister, Peter Weir, has refused to scrap the results arguing that the method was the “least worst solution”.

In a last-minute decision, Welsh authorities told students this week that their A-level grade won’t be lower than the AS grades, which are awarded the year before A-levels.

There has been more uncertainty in England. The education secretary, Gavin Williamson, has faced growing calls to resign, but Prime Minister Boris Johnson has defended the controversial exam results as being “robust” and “dependable”.

Britain’s Exams Chaos: The Fiendish Algorithm Behind Shock A-Level Results

On Friday, Williamson announced a “triple lock” process that would allow students to appeal their moderated results using their teacher’s assessment or results from mock exams taken earlier this year. A third option is to sit an exam in the autumn.

The government has also agreed to cover the cost of schools in England appealing against exam grades they deem unfair.

However, late on Saturday, Ofqual said its policy for mock exam results to be considered as the basis of appeals was “being reviewed” by its board and that further information would come out “in due course”. No reason for the move was immediately available.

Discuss