A Question of Conviction
Halpern, David, New Statesman (1996)
Juries should have the option to find a defendant "probably guilty"
Cases such as those of O J Simpson in the United States and Stephen Lawrence in the UK give rise to widespread misgivings about the criminal justice system. The standard response is to say the evidence was not strong enough to secure convictions. But there is an "unthinkable" alternative, and that is to question aspects of the system itself.
In particular, we should reconsider the untouchable legal concept of "guilt beyond reasonable doubt". Instead of asking juries, and indeed magistrates and judges, to come to an absolute judgment of "guilty" or "not guilty", they could instead be asked to give an estimate of guilt - a judgment of probability.
This information would be used to determine the level of sanction applied. Hence, say, if the jury were 95 per cent sure of guilt for a given offence, the accused might get ten years; but if the jury were only 85 per cent sure, the accused might get three years; and if it were 75 per cent sure, the penalty might be a fine or warning and no custodial sentence.
In such a system, although more people might have punishments applied, the thresholds could be set so the total level of incorrectly applied sanctions would be no higher, and could be lower, than occurs now.
This could be shown formally through the application of a statistically based psychology known as Receiver Operating Characteristic theory. This describes the relationship between false positives and true positives and how their relative proportions are affected by changing the decision criteria. But it can be understood more straightforwardly by thinking about what happens now. The legal system is designed to ensure a high degree of certainty before a person is pronounced guilty, yet errors continue to occur. Innocent people are convicted; while others, though guilty, are not convicted.
This is illustrated - conservatively of course - by Home Office figures. Only 1 per cent of defendants sentenced choose, or are able, to appeal against their convictions. But in 1994, the most recent year for which data are available, more than 5,500 convictions were quashed (320 by the Court of Appeal), a success rate of more than a third.
Most of these errors should be clustered close to the borderline between judgments of guilty and not guilty. By changing the judgment from a black-and-white "yes" or "no" to a grey, graded "maybe", people previously just on the side of guilty (where false positives should be most common) would receive lesser sentences, while those just on the side of not guilty (where false negatives should be most common) would receive minor sanctions. One can argue about how wide this band should be, but the statistical argument of principle is hard to resist.
In mathematical terms it is a similar problem to that in other areas. For example, psychiatrists have historically been keen to make black-and-white judgments about insanity: mad or sane, schizophrenic or manic, and so on. …