Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
58 Risk Perception and Strategic Decision Making: A New Framework for Understanding and Mitigating Biases with Examples Tailored to the Nuclear Power Industry (PSAM-0333)
Download citation file:
- Ris (Zotero)
- Reference Manager
As the economic and environmental impact arguments for increasing the use of nuclear energy for electricity generation and hydrogen production strengthen, it becomes important to better understand human biases, critical thinking skills, and individual specific characteristics that influence decisions made during probabilistic safety assessments (PSAs), decisions regarding nuclear energy among the general public (e.g., trust of risk assessments, acceptance of new plants, etc.), and nuclear energy decisions made by high-level decision makers (e.g., energy policy makers & government regulators). To promote increased understanding and hopefully to improve decision making capacities, this paper provides four key elements. The foundation of these elements builds on decades of research and associated experimental data regarding risk perception and decision making. The first element is a unique taxonomy of twenty-six recognized biases. Examples of biases were generated by reviewing the relevant literature in nuclear safety, cognitive psychology, economics, science education, and neural science (to name a few) and customizing superficial elements of those examples to the nuclear energy domain. The second element is a listing of ten critical thinking skills (with precise definitions) applicable to risk perception and decision making. Third, three brief hypothetical decision making examples are presented and decomposed relative to the unique, decision making bias framework and critical thinking set. The fourth element is a briefly outlined strategy which may enable one to make better decisions in domains that demand careful reflection and strong adherence to the best available data (i.e., avoiding ‘unhelpful biases’ that conflict with proper interpretation of available data). The elements concisely summarized in this paper (and additional elements) are available in detail in an unclassified, unlimited release Sandia National Laboratories report (SAND2005-5730).
The proposed taxonomy of biases contains the headings of normative knowledge, availability, and individual specific biases. Normative knowledge involves a person's skills in combinatorics, probability theory, and statistics. Research has shown that training and experience in these quantitative fields can improve one's ability to accurately determine event likelihoods. Those trained in statistics tend to seek appropriate data sources when assessing the frequency and severity of an event. The availability category of biases includes those which result from the structure of human cognitive machinery. Two examples of biases in the availability category include the anchoring bias and the retrievability bias. The anchoring bias causes a decision maker to bias subsequent values or items toward the first value or item presented to them. The retrievability bias refers to the bias that drives people to believe those values or items which are easier to retrieve from memory are more likely to occur. Individual specific biases include a particular person's values, personality, interests, group identity, and substantive knowledge (i.e., specific domain knowledge related to the decision to be made). Critical thinking skills are also offered as foundational for competent risk perception and decision making as they can mute the impact of undesirable biases, regulate the application of one's knowledge to a decision, and guide information gathering activities. The list of critical thinking skills presented here was originally articulated by the late Arnold B. Arons, a distinguished physicist and esteemed researcher of learning processes. Finally, in addition to borrowing insights from the literature domains mentioned above, the formal decision making approach supported in this paper incorporates methods used in multi-attribute utility theory.