Algorithms are no substitute for professional judgement: No. 10 needs to take note

Posted on

The English A level (and no doubt, GCSE) exam fiasco is unfolding, just as it has in Scotland.

Scotland has had to abandon its exam algorithm because it has been widely considered to be unfair. Teacher forecasts have been substituted.

In anticipation of the same crisis in England, Gavin Williamson has announced that students might use their mock A-level results instead of those to be allocated tomorrow.

I suggest that, as usual, in what is an embarrassment for both governments Scotland has been substantially more successful in addressing the issue than England has been, so far.

I should admit that I have personal interest in this issue: my younger son is expecting results tomorrow morning, and success or otherwise in securing a university place depends upon them. I fully accept that my objectivity is compromised as a result. But, that said, the issue I want to raise is a systemic one. The difference in approach between England and Scotland is not down to reaction to last-minute panic. It is about something more fundamental, and that is the confidence that these administrations display in the professional people who serve their communities.

The only people able to appraise the likely results of those set to take these exams in the absence of an examination were those teachers who had got to know the pupils in question, who had appraised their work, and who had as a consequence the opportunity to create a professional opinion upon it, which opinion could be second checked by another teacher familiar with the system in which they were working. Logically, teacher assessment was the only way in which this issue could be resolved.

Of course there were issues to address. Some schools would be under pressure to overstate likely results. Some teachers might have unwittingly understated performance for some social groups: this is known to be a problem. And exceptional pupils were always going to stand out from the crowd. But, I suggest, all these issues could have been addressed, and in a much better way than has been done.

First, it should have been stated from the outset that teacher assessment was what mattered: after all, like it or not that is now happening.

Secondly. therefore, those assessments should have been provisionally published when submitted, months ago.

Third, OfQual should have stated in advance when they would address issues at a school, in particular, for example because of aberrational increases or decreases in performance. It should have been made clear which schools were being investigated. Evidence from the school, parents and pupils should have been sought. This would have involved a minority of cases, of course. And it would have been fair.

And fourth, pupils should have been allowed to appeal their grades three months ago. That way time would have been available to look into cases and top resolve issues to ensure satisfaction with the outcomes.

This would have been fair to all - and a disincentive to school game playing as well. It would also have respected professional opinion, whilst allowing that it can be wrong (as it is every year when exams are regarded on remarking).

This would have worked. But it would have required respect for the opinion of teachers as professional people. And the government, at least in England did not, and still does not, have that. Instead, it believed that everyone was out to cheat. That was a false, crude and deeply patronising assumption that says more about that who held that view than it does about teachers. And because the professional opinions of those who know have been ignored we have a mess. As we have had on either issues of as much significance of late.

I am not saying all professional people are virtuous. I am certainly not saying that they cannot make mistakes. But I am saying algorithms are no substitute for professional judgement. And I do think it's time that the government learned that this is the case, and that they do not necessarily know best.

Very specifically, it is always true that there are occasions when knowledge of statistical distributions is no substitute for sound judgement. Some in No. 10 need to take note.