Research: People do not allocate enough resources to risks with lower probability of survival

[Study authors: Adam Elga, Jian-Qiao Zhu, Thomas L. Griffiths]

Summary: When asked to divide a fixed budget to fight several independent existential risks, experimental participants did not allocate enough resources to the risks with lowest probability of survival.

Excerpts:

if we face catastrophes with survival probabilities of 4% and 40%, it is more valuable to increase the probability of surviving the first catastrophe from 4% to 9% than it is to increase the probability of surviving the second catastrophe from 40% to 80% (since we care about the product of the survival probabilities and 9% ⋅ 40% > 4% ⋅ 80%). This dependence of optimal allocations on baseline chances of survival is a distinctive and somewhat counterintuitive feature of situations in which incentives are multiplicative (Lewis et al., 2023; Lewis & Simmons, 2020). [Reference added 2025-08-24: Ord 2020, Appendix D]

But

on average people allocated resources more evenly among the risks than was optimal.

Figure legend: The optimal model (blue dots) compared with the mean normalized allocations made by participants (grey dots). Each vertex represents full investment in a given risk. Connecting lines indicate correspondence between model predictions and empirical data.

Conclusion excerpt (emphasis added):

our results suggest that an extremely natural and seemingly obvious practice – presenting risk mitigation options individually for comparison – is suboptimal. Put bluntly: while people are relatively good at choosing individual items for their grocery carts given each item’s price per ounce (a linear allocation problem), people are less good at the nonlinear allocation problem of choosing individual existential risk interventions based on each intervention’s risk reduction per dollar spent. As a result, it may be better to instead present decision makers with competing total allocation profiles (specifying how much is to be invested in each risk), where each profile is explicitly labeled with its estimated reduction in overall existential risk (the risk that any catastrophe occurs).