Sunday, March 27, 2011

Risk Communication - Perception and Deception

Another simple example of poor or misleading risk communication can be found in the O. J. Simpson murder trial.  One piece of information that OJ’s defense team were able to quash was the prosecutions assertion that spousal abuse leads to murder.  The defense argued that Simpsons history of assaulting his wife, Nicole Brown Simpson was not relevant to whether or not he had murdered her.

Alan Dershowitz, a Harvard Law Professor in his book about the case argued that in the United States:
As many as 4 million women are battered annually by husbands and boyfriends. Yet in 1992, according to the FBI Uniform Crime Reports, a total of 913 women were killed by their husbands and 519 were killed by their boyfriends.  In other words, while there were 2 ½ to 4 million incidents of abuse, there were only 1,432 homicides.  Some of these homicides may have occurred after a history of abuse but obviously most abuse, presumably even more serious abuse, does not end in murder” [8]

Essentially the defense argued that based on these figures, there is less than one homicide per 2,500 incidents of abuse and they used this to argue that there was no evidence of domestic violence being a prelude to murder.  While this is factually true, it is not a useful statistic.

Not only was it not useful but it may well have misled the court. The correct question to ask should have been: “How many women were murdered by men who had previously abused them?”  At the time of the trial, statistics showed that out of every 100,000 battered women, 45 were murdered. Of those 45, 40 were murdered by men who had previously battered them.  In short, 90% of murdered women who had been battered by their partners had in fact, been killed by their partners.  Rather than a 1 in 2,500 probability, past data suggested a statistically significant probability of 90% that OJ was the murderer.

It’s also worth bearing in mind that the death of a woman at the hands of a partner who has previously battered her may appear predictable in hindsight but when only 1 in 2,500 battered women go on to be murdered, this statistic has little if any utility when predicting the likely risk of murder.
Does a 90% probability constitute evidence of OJ’s guilt? Of course not!  Whether or not it would have influenced the jury is another story. We can never say for certain but ask yourself – is it likely that the way in which this information was presented would have influenced your views?

A Call to Action

Given what you now know, is it any wonder that our political leaders and the general public have trouble understanding and prioritising risks such as terrorism, crime, health, national security and hundreds of other risks.  It seems that even in the 21st century with all our amazing communications technologies we have a long way to go to master the simple act of communication risk in any meaningful fashion.  The groundwork on how to present risks using natural frequencies has been done for us by practitioners in areas such as medicine, psychology and statistics. Perhaps it is time that we as risk professionals, managers and policy makers started to look more closely at exactly how we choose to present our risk data?

[8] Dershowitz, Alan (1997), Reasonable Doubts: The Criminal Justice System and The O.J. Simpson Case, Touchstone, New York, USA.

No comments:

Post a Comment