Four years ago, an American student fooled a computerised essay-marking system into commending him - for a nonsensical essay which claimed that rabid aardvarks posed a threat to children’s outdoor activity centres.
We revel in such coups, because they prove that “warm bodies” - the fetching term that designers of artificial-intelligence systems use to describe human beings - can still outperform machines in tasks that require real intelligence and insight rather than flashy computational skills.
But can we really? This week’s report explaining how variable human exam-markers can be (page 5) is a reminder of our fallibility. But the Intelligent Essay Assessor (page 7) never gets bored, sleepy or forgetful. It might have failed Ernest Hemingway for writing sentences without subjects, but we should not underestimate its “intelligence”. The assessor studies the combination of words in a student’s written answer and computes its similarity to model essays which it has stashed away in its electronic brain. It’s as simple - and complex - as that.
Before long, teachers may accept that such a system is as important a labour-saving device as the dishwasher; but if that day comes we’ll feel a little sad and deflated all the same.
