Very interesting presentation by David Spiegelhalter of Cambridge University on “Perceptions of Risk – Communicating Risks and Deeper Uncertainties In Words, Numbers & Pictures“. David is the Winton Professor for the Public Understanding of Risk, Department of Mathematics at Cambridge University and is involved with the website understandinguncertainty.org.

David started by saying that when communicating risk what is needed is a user-friendly, easy to understand unit of risk. One example he gave was a one in a million change of death, which he termed a “micromort“. He said that on average around 50 people a day die of unnatural causes in England and Wales every year, so this means that with a population of 50 million a person’s exposure in England and Wales is 1 micromort. He then compared various means of transport and the distance that would need to be travelled to reach a 1 micromort measure:

• Walking – 12 miles
• Cycling – 20 miles
• Car – 217 miles
• Motor Bike – 6 miles

So I guess those of you with motor bikes should take note above!

He emphasised that also peoples reaction to risk is very interesting, and gave the example of a UK health official recently sacked for suggesting that the hobby (addiction?) to horse riding (he termed it “equesy”) was just as risky as taking the drug ecstasy – statistically what the official said stacked up but it was simply not culturally acceptable to suggest such an association. No great advertisement for the NHS, but there are around 3753 deaths a year with an average of around 135,000 people in hospital at any one time. This works out at around 75 micromorts if you are in hospital which is around twice the level faced by troops in Afghanistan!

Obviously the above has a number of biases, but David was trying to illustrate how to compare risks and how people are not used to assessing them objectively. In particular, given a choice between probabilities of 1 in 10, 1 in 100 and 1 in 1000, around a quarter of the public would choose 1 in 1000 as the highest probability given it contains the “highest” number.

Whilst the above refers to the denominator with a constant numerator, given the choice between drawing from a bowl containing 1 sweet and 8 marbles and another bowl containing 5 sweets and 45 marbles, 53% of people choose the one containing 5 sweets (because it contains “the most” chances of getting a sweet).

David went on to test the audience on a few trivia questions using what he termed a “quadratic scoring” scale that asked the participant to select a multi-choice answer but to also associate a level of confidence with it. If right and confident the marks given would be high, but if wrong and confident the penalty mark would be much larger. He said that such scoring often produced interesting results and changed people’s views, often with young men doing worse (testosterone not being good for risk seemingly!).

He showed how probability density seemed to better understood if represented by density of ink rather than the usual bell curves etc. He suggested that results should come with some more warning of how reliable they are to stop simple acceptance of the numbers reported as “truth”. He described how risk is measurable whereas uncertainty is not, which led to the inevitable references to the wisdom (?) of Donald Rumsfeld.

Good fun talk with some great points to make – how humans (and bank management boards?) understand risk is interesting and to some extent surprising (see earlier post for a different slant on human perception of maths). Obviously it is accepted now that simple VAR measures are not enough, but even with the move towards scenario based methods then how to produce a simple but meaningful summary of risk for management is still challenging.