Nuclear Power Risks and the Optimism Bias

Psychology professor Tali Shalot writes in The Optimism Bias that 80% of humanity has an irrational positive bias toward themselves and their families. It is likely to be an adaptation that helped us survive because she finds that this bias increases the probability that we will obtain our goals. In contrast, a bias toward pessimism is associated with depression.

Most people rate themselves as above average in several traits such as driving ability, appearance, intelligence, likability and, of course, modesty. They think their own children are above average in many ways as well. Newlyweds know that 40% of marriages end in divorce, but they think that theirs will not. People have these biases toward themselves even if they are pessimistic about other people or society in general. Life would grind to a halt without this bias. This optimism bias is advantageous in maintaining emotional resilience and obtaining goals, but there are definitely situations, especially for people with heavy responsibilities in leading nations or managing dangerous technology, in which this bias can be disastrous. Evolution did not prepare us to be nuclear plant managers, so our built-in cognitive wiring for risk assessment is unreliable and must in many circumstances be over-ruled by the power of reason.

The Japanese nuclear industry and its regulators were definitely blinded by complacency, fear of career setbacks, fear of financial losses, and optimism bias in the years leading up to the Fukushima Daiichi catastrophe. Everyone knew that recent information about seismic and tsunami risks required a reassessment of the original assumptions about the safety of nuclear plants, but decisions about the necessary upgrades were put off in the hopeful belief that the problems would be solved before disaster struck. They erroneously thought that it probably wouldn't happen in the near future, but a rational study of the probabilities reveals the counter-intuitive finding that the farther you look into the future, the lower the probability of an event occurring.

In The Better Angels of our Nature, (p.202-203) Steven Pinker discusses the inability 1,000 research subjects to correctly understand the probability of an event happening in the future. In the problem discussed, the probability of lightning striking a home is given as once a month. People can understand that the probability of lightning striking the home tomorrow is 0.03 (1/30 days of the month). However, when subjects were asked to state the probability of lightning striking 2, 3 or 4 days into the future, most people said the probability was the same or higher. They were unable to see that there was a decreasing likelihood of many days passing without a lightning strike. The probability goes down the farther you consider into the future. The lowest probability is that 29 days will pass without a lightning strike and lightning will hit on the 30th day from now. Of course, when you wake up tomorrow it is a new “today,” so the calculation resets, but from the perspective of today, tomorrow is the day with the highest probability of a lightning strike. 
The probabilities are perhaps easier to understand with the last day of the month as the reference point for the present. If 29 days have passed without a lightning strike, the likelihood of a lightning strike today is 100%.These insights should tell safety engineers that the time to act on mitigating a risk is right now. The greatest error is to kick the can down the road on the assumption that it probably won’t happen anytime soon. Other common mistakes are to think that the risk assessment is flawed and to hope that new research done with favorable methodology will provide a reason for inaction and not spending money on the problem.

The optimism bias messes with our estimations of probability and makes us think highly unlikely bad things won’t happen, but think highly unlikely good things will happen. If every year there is a 1/1,000 chance that a mega-tsunami will destroy the cooling system of a nuclear power plant, the operator concludes that there is no urgency because it probably won’t happen this year. On the other hand, if the workers at the power plant see advertisements that the state lottery jackpot is getting very large, they will buy more lottery tickets than usual. They know that millions of people are doing the same thing, more tickets are being sold, and the probability of winning is becoming exceedingly small, but they buy tickets anyway. These are two low-probability, high-impact scenarios, but the irrational decisions about them differ because the impact of the former is unfavorable and the impact of the latter is favorable.

Finally, from a certain perspective, probabilities are meaningless numbers. Gamblers don’t care about the odds stacked against them as long as they might win the next round. When risk increases by factors of 10, at what point does the risk become unacceptable? Whether the probability is 1/100, 1/1,000 or 1/10,000, the impact of the unwanted event may be nasty enough to render the difference meaningless. Regardless of the probabilities of the event occurring, the only thing that matters is that it can happen tomorrow.

So now think about the present and future risk to nuclear power plants, based on lessons learned from the Fukushima catastrophe. A whistleblower at the U.S. Nuclear Regulatory Commission recently spoke to journalists regarding his concerns about the Oconee Nuclear Power Plant in South Carolina being destroyed by a dam failure:

"The probability of Jocassee Dam catastrophically failing is hundreds of times greater than a 51-foot wall of water hitting Fukushima Daiichi. And, like the tsunami in Japan, the man-made 'tsunami' resulting from the failure of the Jocassee Dam will - with absolute certainty - result in the failure of three reactor plants along with their containment structures. Although it is not a given that Jocassee Dam will fail in the next 20 years, it is a given that if it does fail, the three reactor plants will melt down and release their radionuclides into the environment."

In hindsight we can all say that TEPCO or Japanese regulators should have shut down Fukushima Daiichi and built better seawalls and backup power systems, but if the probability of dam failure in South Carolina is higher than the tsunami risk was at Fukushima, it follows that the correct foresight regarding the Oconee plant is that it should be shut down immediately until the risk can be eliminated. The dam could fail tomorrow. The same conclusion could be made about the risk posed by earthquakes, tsunamis, floods, hurricanes and tornadoes at many other nuclear plants. The nuclear industry has spoken a lot about lessons learned from the Fukushima nightmare, but if another catastrophe occurs because of a risk ignored, the public will conclude that nothing was learned.

Further reading:

Environment News Service 23 Nuclear Power Plants at High Risk of Tsunami. September 24, 2012