I’m a engineer by training, and I was taught that:

risk = probability of something going wrong x resulting impact of that event

There’s a big problem with this equation – it is out of sync with the public perception of risk. Our gut instinct is to fear one big accident much more than lots of little ones. So an air crash leading to the deaths of 100 people will get much more press coverage than 1000 car accidents killing one person in each accident.
There is a tendency to consider such risk perception as completely irrational, but in practice it provides an essential buffer to a number of metarisks (risks in measuring risks):
1. Mistakes – either mistakes in calculations, estimations and measurements, or not considering all the factors that apply. Such mistakes usually have much more serious consequences when working with high impact risks than in the case of lots of low impact risks.
2. Deliberate distortion of results – history is littered with cases of risks being deliberately played down in order to allow a project to proceed.
3. We tend to underestimate risks we are familiar with, or overestimate risks that we are unfamiliar with. Therefore people “in the know” tend to subconsciously play down risks.
So the public perception of fearing big impact risks balances the tendency of risk managers to make mistakes, be blasé or even, in the worst instances, distort results. The Gulf of Mexico oil spill was judged to have a negligible risk of the catastrophe occurring and a negligible impact – the latter prediction was certainly incorrect and the former is very likely to have been incorrect. Yet if you asked a member of the public whether drilling for oil in mile deep water is “risky”, most would say “sounds like it to me”.
Maybe we should trust our gut instincts a little more and our calculations a little less. But at the very least we should understand why the public perception of risks is as it is and use it as a spur to make sure that a ‘rational’ approach to risk management is not the result of wishful thinking.