The Limits of Prediction

Abstract

Were I to talk about the circumstance which seems to me more than any other to limit prediction in Higher Education— both the prediction of individual success and the estimate of total numbers capable of a particular level of academic work— I should discuss, not the intelligence of candidates, but the foibles of examiners. Our forecasts are validated against a criterion which has been shown again and again to be unreliable, yet each of us seems able to persuade himself that it is other examiners who go astray, and in general that with care and patience we can make the present system work well enough. This view is not unreasonable when we have in mind the random lapses of attention and changes of mood to which we are all subject—what I have called the ‘noise’ in the system²—but it is not true of adaptation. Human beings are incredibly adaptable to slow consistent changes of any kind. As a result what they respond to as neutral or normal or average, within quite wide limits, is not anchored in the outside world, but varies with their own past experience, especially their recent past experience. This can produce ‘drift’, a steady change in standards over time.