Note: this page links to an outside resource that provides a significant amount of the coverage that a GiveWell shallow investigation would aim for. Rather than repeat the outside researcher’s process, we link to their work here and offer some reflections.
Process
In consultation with GiveWell staff, Nick Beckstead (who was not then a GiveWell employee, but has since become one) decided to conduct a shallow investigation of improving disaster shelters to increase the likelihood of recovery from a global catastrophe. We paid him for his time on the project.
Reflections
Nick’s shallow investigation reaches a relatively pessimistic conclusion about the potential returns to philanthropic effects to increase the likelihood of recovery from a global catastrophe by improving disaster shelters. We find that conclusion persuasive, for a couple reasons:
- There appears to already be a fairly high level of shelter protection globally. Nick documents the existence of continuity of government bunkers in the United States and several other countries, and was told that further information would likely be hard to obtain because of government secrecy around these issues. In addition, one of the major private shelter providers claimed to have built more than 1,000 shelters that are robust to many kinds of disasters, with room for more than 10,000 people. Another company has built a shelter in Indiana that it claims can support 1,000 people. Nick also points to the number of people who are on submarines at any one time and the many people who live in highly remote locations with little outside contact as evidence that the current level of robustness to the kinds of threats that bunkers might help protect against is already fairly high.
- The set of global catastrophic risks for which an improved shelter capacity might conceptually be helpful seems to be quite limited. For bunkers to be helpful in preserving the long-term survival of humanity, it seems that they would need to be robust enough to save people from a threat that would otherwise kill everyone, but threats that might kill everyone except a few people in a bunker are relatively difficult to imagine. Most threats that might conceivably lead to human extinction would be unlikely to be stopped by a bunker, and other potential global catastrophic risks appear extremely unlikely to cause human extinction. The only potential risk that Nick comes up with for which shelters seems fairly likely to be helpful in preserving the long-term survival of humanity is an engineered pandemic, but there seem to be better ways to confront that threat at this point than through additional shelters.
Based on these considerations, we agree with Nick’s conclusion that research on other approaches to dealing with global catastrophic risks is likely to be more promising at this point.