Atomic Horizons: Forecasting Nuclear Risks

Life inherently involves risk. Every decision carries some degree of it. For instance, when you leave home to get groceries, you might encounter a traffic accident or even face assault. Each day, there’s a small but increasing likelihood of developing cancer or heart disease, which are the leading causes of death worldwide. These risks accumulate over time, and eventually, “your luck will run out.” 

However, we can influence these risks through our choices. For example, you can walk on the pavement instead of the street and look both ways before crossing. You can opt to live in a low-crime area, eat healthily, and limit alcohol consumption. While your luck will still eventually run out, these choices may delay that day and lessen the severity of the consequences. 

Just as everyday life involves unavoidable risks, the existence of nuclear weapons creates an inherent risk of their use, whether intentional or accidental. There is always a non-zero probability of nuclear conflict or accidents involving nuclear materials. This risk varies across regions, with some areas more likely to experience nuclear conflict than others.  

As long as nuclear weapons exist, the cumulative risk of their use increases over time, even if the annual risk remains low. We intuitively understand this in our daily life. Say that you are a frequent flyer, and you have a 1-in-100 chance of experiencing a significant flight delay on any given flight. If you take 100 flights over one year, it is much more likely than not that you will experience at least one significant delay. 

Nuclear risks are not exempt from this law of large numbers.  A forthcoming study by the Forecasting Research Institute, supported by the Open Nuclear Network*, estimates the annual risk at approximately 0.092% based on a global elicitation of 117 experts and 43 superforecasters. While this number might seem reassuring, it translates to a 1-in-20 chance of a significant nuclear exchange (resulting in more than 10 million dead) by 2100. Although this may seem distant and the probabilities are low, there is a notable chance that several children born today will witness a nuclear detonation in their lifetime. 

You may now think that the risk seems exaggerated. This is a natural reaction, known as the salience bias. It leads individuals to overestimate the importance or likelihood of events based on their prominence. This bias is common in gambling, where lottery advertisements highlight big winners, making these events seem more frequent than they are and encouraging more people to buy tickets. Despite the odds of winning a significant prize often being around 1-in-50 million, the salience of the winners distorts perception. Moreover, as humans, we tend to exaggerate risks that provoke a strong emotional response (such as terrorism or crime) and underestimate risks that do not elicit the same response (such as abstract risks like climate change or a nuclear explosion in your city).  

The risk is higher than it seems, but we can influence risks through the choices we make. Just as individuals can make decisions to reduce personal risks, nations and international bodies can take actions to reduce the likelihood of nuclear conflict or accidents. Over time, a minor course adjustment can significantly impact cumulative probabilities. If we manage to reduce the annual probability to 0.04%, the likelihood of occurrence by 2100 drops to 1-in-33. However, such adjustments can also increase risk. If nuclear weapons become normalized, and the annual probability rises to 1%, we’re more likely than not to see their use by the end of the century. At a 5% annual probability, nuclear weapons use is almost certain before the century ends. 

We can also take steps to reduce the severity of consequences through our choices. Disaster risk is commonly defined as a function of hazard, exposure, and vulnerability. Given that the use of nuclear weapons is an anthropogenic hazard, we can directly influence the probability of it occurring by the choices we make to reduce the risk. We can also reduce risks by ensuring that nuclear arsenals are kept as small as possible (thus physically limiting the damage they can cause) and that nuclear-armed states, in their doctrines and targeting, expressly avoid the use of these weapons against major civilian centers or other unlawful targets. Ultimately, the only way to eliminate the risk is to remove all nuclear weapons from national arsenals and safeguard against their return. As the Canberra Commission on the Elimination of Nuclear Weapons wrote in 1996: “the proposition that nuclear weapons can be retained in perpetuity and never used – accidentally or by decision – defies credibility.” They further observed that “the only complete defense is the elimination of nuclear weapons and assurance that they will never be produced again. 

Thinking about nuclear risks in terms of probabilities allows for a more precise understanding of potential threats. When I first began exploring this field, a diplomat told me it was unnecessary, as “we all understand the risks.” I don’t think we do. Instead, my sense is that we don’t have a clear picture of what the risks associated with nuclear weapons are or how these risks could escalate or evolve. The diplomat I spoke to was not dismissive; his comments stem from a reasonable position. Extremely low-probability, high-consequence disasters, such as nuclear use, are exceptionally difficult to forecast with precision. However, this does not mean that the process lacks utility. As James Acton points out in his article  How I Learned to Stop Worrying and (Sort Of) Love Nuclear Forecasting: 

“The process of forecasting—identifying sequences of events that could lead to nuclear use and assigning probabilities—would provide a systematic way to allow policymakers to identify each other’s assumptions and hence understand why they disagree. Such understanding, in turn, would allow them, for example, to task the intelligence community more effectively, ultimately improving the quality of the advice they provide to a president.” 

Furthermore, similar to the approach taken by the International Panel on Climate Change when assessing the future of our climate, we need to adopt a long-term perspective in nuclear risk assessment. While the probability of nuclear use might be low in any given year, the cumulative risk over decades becomes significant. Extrapolating risk over time makes future scenarios more tangible, especially since we all intuitively understand that continuing to make hazardous choices increases the likelihood of our luck running out sooner rather than later.

Moreover, Christopher Ford has argued that nuclear weapons can be used, and indeed are used, to enable and provide cover for conventional aggression. These policies are designed to confront intervening powers with an unappealing dilemma: either allow the aggression to succeed or risk initiating a wider war, potentially escalating to nuclear exchanges. If this argument is taken to its logical conclusion, it’s easy to envision a world where revisionist authoritarian states expand under the cover of their nuclear umbrella, subjugating peoples, and their liberty, as they proceed. Such a future would be bleak, and while we might avoid outright catastrophe, it’s hard to argue that the generation living in the year 2100—who already walk among us—would be fundamentally better off. The great paradox is that nuclear weapons, under such conditions, would enable the very future that the scientists of the Manhattan Project sought to avoid with their creation. 

The splitting of the atom did not end the world, but it irrevocably altered our understanding of it. It ushered in a new era, transforming science, medicine, energy, and industry. Yet, this profound discovery also cast us under the shadow of catastrophic risk. The true measure of our progress lies in our ability to wield this power responsibly. We must understand and internalize these risks with precision if we aspire to shape a positive future. The choices we make today will define not only our era but also the very future of our world. 

Andreas Persbo is the Director of Open Nuclear Network (ONN), a programme of PAX sapiens. Andreas is a recognised expert in arms control and its verification, and in particular nuclear non-proliferation and disarmament. Over the last few years, he has also sought to expand his knowledge of great-power relations, specifically West-Russia dynamics. He is the author of some 60 book chapters, research reports and briefing papers and has delivered lectures and presentations at several dozen conferences and workshops.

* The Forecasting Research Institute (FRI) and Open Nuclear Network (ONN), in association with the University of Pennsylvania, are conducting a comprehensive study aimed at understanding the spectrum of expert opinions on the risk of nuclear catastrophes in the upcoming decades. This includes collecting views on the overall likelihood of such a catastrophe, the most likely risk pathways and factors influencing risk, and the potential impact of various policy choices. This initiative seeks to both quantitatively and qualitatively capture the views within the nuclear policy community and to pinpoint key areas of contention that can be empirically assessed.    

To date, we surveyed 117 nuclear experts and 43 superforecasters and are ready to share some findings.   

To present the interim findings and open it up for a broader discussion, ONN and FRI are organizing a side event “A Gamble of Our Own Choosing: Forecasting Nuclear Risks” at the upcoming 2024 NPT PrepCom in Geneva, Switzerland.  

[Learn more and register]