28. Ignoring a risk because of its insignificance according to an expert

This ties into the above point. If an expert thinks a risk is insignificant, he may be wrong. It is also necessary for the expert to quantify the precise degree of insignificance they are talking about, which they often refuse to do or are unable to do. For instance, say that we determine that the probability that a certain particle accelerator experiment destroys the planet is one in a million. That sounds low, but what if they are running the same experiment a hundred times a day? After a year, the probability of doom is already 1/300. After 500 or so years, the probability of doom approaches unity. So, “one in a million” may actually be a quite significant risk if the risk is repetitive enough.

Aside from having a proper understanding of insignificance, we also ought to keep in context the relationship between an estimate of insignificance and the chance that the expert making the prediction is in error. Say an expert says that an event only has a one in a billion chance, but there is a 50% probability that they are completely wrong. In that case, the real insignificance might be just one in two, or one in twenty. The probability that an expert is just plainly wrong often throws off the estimate, unless there is a rigorous statistical or empirical basis to confirm the estimate. Even then, the empirical data may lie, or there may be an error in the statistical calculations, or a mistaken prior.