What I am arguing is that low-probability events can be treated as effectively certain over large enough sample sizes. Similarly, dangerous jobs impose a socie

What I am arguing is that low-probability events can be treated as effectively certain over large enough sample sizes. Similarly, dangerous jobs impose a societal health burden via workplace accidents. Each accident is one-off but in aggregate, a non-zero accident rate is a certain outcome. Containment breaches of contagious diseases, while each a one-off event, are effectively certain over large enough sample sizes- and therefore should be considered as part of the cost of doing research into them.