Bayesian analysis of public views on the safety of nuclear developments

Abstract

The sodium leak accident at Monju, a prototype fast breeder reactor that the Power Reactor and Nuclear Fuel Development Corporation (PNC) has operated since 1995 in Japan, in combination with the PNC's management of the accident, has created public distrust not only in the safety of Monju, but also in the nuclear development policy of Japan. In order to learn a lesson from these events and to assess how a relationship of mutual trust can be established with the public, this paper simulates the processes of establishing public confidence in reliability of a nuclear plant in cases of accidents, no accident, a cover-up, and so on. In this paper, public confidence is defined as the public evaluation of cumulative probability under a certain level of accident rate, conditional on the information available to the public. The conditional probability is estimated by use of Bayes' Theorem. The simulation shows that (1) public confidence is lost by only one accident in an early stage of operation, and can then be recovered only by many subsequent years of accident-free operation, but never by a cover-up, and (2) the more information that is provided to the public, the better the relationship of mutual trust that will be established, especially at an early stage of plant operations.