Why we should use algorithms more in education

At this time of year I always thank my lucky stars that I am not an examinations officer. The potential for getting things wrong by one small administrative slip-up is huge. I've been in examination halls where some candidates have calculators, some don't; some have dictionaries, some don't; some need all their scripts presented to them on green paper, whilst still others are entitled to extra time. The extra time entitlement, of course, is not the same across the board: some get 10%, some 25%, some 50%, some even, on occasion, 100%. Complexity rules. The potential for honest mistakes is huge.

In my opinion, across the country, too many children are getting special treatment of some sort; the playing field needs to be re-levelled. It concerns me that sharp-elbowed middle-class parents game the system, though given that the system is so gameable they can hardly be blamed. They only want the best for their children. I know of selective independent schools where upwards of 25% of pupils get some dispensation or other in the completion of their exams - this is madness and patently not fair. The silent majority, who quietly get on and accept their lot, are being done over.

But I wonder whether technology could come to the rescue? I should explain, before I begin, that I have a new found faith in algorithms. Algorithms make better decisions than humans in the majority of cases, even - perhaps especially - in cases where human judgement or a degree of emotional intelligence is seen as essential. Algorithms don't make mistakes, they are not swayed by first impressions, nor do they get tired, hungry or irritable. As you can see, I'm a fan. If you're not yet convinced you should read Thinking Fast and Slow.

Firstly it can't be long before the majority of exams are done on laptops rather than on paper. Administered in this way, lines of computer code rather than humans could do many of the routine tasks involved in administering an examination. In the exam halls of the future I imagine papers going live on the screens of candidates at the allotted time (no more: 'You may now turn your papers over'). The machines will also close scripts automatically and make allowances for any given candidate's extra-time allowance (or screen colour, or dictionary on/off setting etc.) As I regularly tell my pupils: 'Computers don't make mistakes, programmers do'. Instances of human error compromising someone's life chances will almost vanish.

Secondly, examining in this way would allow for a richer set of data to be produced on the abilities of the candidates themselves. It seems to me entirely fair that future employers and universities ought to know the length of time completion of the paper took, and whether additional aids were used. However much the beneficiaries of extra time might protest an A* with 50% extra time is not the same as an A* in normal time. Added to which, we are constantly told that employers and universities find it difficult to discriminate between candidates. Imagine what they might do with the richer data set that online exams would afford (time, speed, breaks, corrections, even eye tracking..?)

More than all of this though algorithms should be more widely used in the assessment of children's abilities long before they've ever set foot in an exam hall. It has been convincingly shown that professionals routinely make duff judgments in the very fields they purport to be most knowledgeable about. And I have railed against the industry that preys on middle-class sensibilities resulting in over-diagnosis of special educational needs. Algorithms should be brought to bear, largely replacing expensive educational psychologists for most of the routine applications for extra time.

Algorithms don't feel moral obligation, are ruthlessly impartial, and do not depend on large consultancy fees. We should trust them more than we do.

comments powered by Disqus