Which risks should we study? Which methods should we use? Who should we be impacting? How should we communicate? What needs to change?
On 1st July 2021, CSER presented the first of its summer panels, "Doom and Doubt: Uncertain Futures & Open Questions about Existential Risk" showcasing three CSER researchers from very different backgrounds. The panel discussed how we can have frank, open, and vulnerable conversations about existential risks that explore our doubts, uncertainties, and confusions, without detracting from their urgency or importance. Bringing together insights from science, law and philosophy and years of studying different kinds of risk, the panel offered challenging insights into how to live up to the highest epistemic standards while also confronting urgent threats.
As well as the discussions recorded in this video the event also saw a lively debate between attendees. Some questions that came out from this discussion that we were not able to respond to during the event itself include the following:
- As the recent pandemic has shown, politicians & policy-makers are not very good at proactively preparing for uncertain near & long-term future events. How would the panellists recommend tackling this problem for existential risk preparation?
There are many lessons that we can learn from the pandemic. What we have experienced is certainly a long way from the best-case scenario for responding to a global-scale disaster, but it is also a long way from the worst-case scenario too. Rushing to draw conclusions for what we can take from this experience in preparing for other risks, which could be quite different in severity and nature, is unlikely to be helpful. CSER is currently in the early stages of a multi-year project to fully engage with COVID-19 and the lessons we should learn from it.
- There are methods for dealing with this sort of "decision-making under deep uncertainty." Has CSER or any of the panellists explicitly tried to apply these or similar approaches to existential risks?
Our work has engaged with a number of these approaches. For instance, SJ Beard (and colleagues) advocate for the use of Robust Decision Making in their recent paper "Assessing Climate Change's Contribution to Global Catastrophic Risk". Luke Kemp has also applied adaptive policymaking including in his report "The Cartography of Global Catastrophic Goverenence" and further work on these themes is forthcoming. However, our primary aim as researchers is to understand risks and mitigation strategies better, while these tools are most useful for decision-makers who use our research.
- I've often thought that a version of Pascale's Wager could be applied fruitfully to the subject of existential risks; any thoughts on this?
There is a long history of considering Pascal's wager, and its counterpart Pascal's mugging, in the context of existential risk. At a theoretical level, it is very easy to see cases in which the sheer value of humanity's future might lead to Pascal's wager-type arguments, where even trivial actions that might help to secure that future become necessary even though they appear somewhat irrational. However, it is important to stress that these theoretical cases do not conform with actual practical decisions we face about existential risk, where the actions that would help to secure humanity's future tend to be lower cost and more intuitively plausible, such as doing basic research into AI safety before developing the technology or enhancing security around novel applications of biotechnology, while others are likely to be net beneficial as well as reducing the level of global catastrophic risk, such as tackling climate change and the biodiversity crisis.
- If climate science is as settled as the panellists suggest then why were no climate modellers able to predict the slowdown in global warming from 1998 to 2012, even though they could explain it afterwards, and what does this tell us about our ability to model future risks?
Climate science is settled in the sense that we understand which factors are likely to produce changes in the climate and how these operate, that does not mean that there are not many things about which we are uncertain. One of these is the exact level of the climate response to changes in the concentration of greenhouse gasses due to the many feedback loops involved. Another is about the prediction of phenomena that are more granular than climate, such as predicting the weather. Climate scientists tend to model the climate in 30-year time blocks, within which the 1998-2012 anomaly does not appear as it is less than 15 years in duration. So it's not clear that climate scientists even 'should' have been trying to predict it, let alone it being a problem for the science that they did not. Nevertheless, it is important to note that while the long-term trajectory of climate change is a significant driver of global catastrophic risk, such short-term phenomena may well be even more significant in their impacts precisely because they are so unpredictable and potentially extreme.