Making the Most of Limited Evidence In Global Catastrophic Risk Analysis (CCCR2018)

Published on 17 April 2019

Seth Baum is Co-Founder and Executive Director of the Global Catastrophic Risk Institute, a US non-profit think tank.

Global catastrophic risks and existential risks are characterized by a fundamental limitation of evidence due to the fact that global human civilization has never previously been destroyed. Whereas traditional risk analysis uses empirical datasets on a given type of harm, this is not an option for these risks. Instead, analysis must work from a more limited evidence set.

Dr Baum surveys the types of evidence available and how they can be used in risk analysis. Evidence types include precursor events, analogous events, general processes, and expert judgment. Risk models can integrate this evidence and use sensitivity analysis and related techniques when evidence is missing. Risk theories of zero-fail data can provide further insight. Additionally, a focus on the informational needs of decision-making can be of further value because detailed risk calculations are often not needed to make decisions. These points are illustrated using examples from nuclear war and artificial general intelligence.

This talk was given at 2018’s Cambridge Conference on Catastrophic Risk (CCCR2018), the Centre for the Study of Existential Risk’s major international conference, supported by the Templeton World Charity Foundation. It focused on four challenges faced by research communities focused on existential and global catastrophic risk research: Challenges of Evaluation and Impact; Challenges of Evidence; Challenges of Scope and Focus; and Challenges in Communication.

Subscribe to our mailing list to get our latest updates