This paper tests a different way of choosing climate policy when models disagree. Instead of picking the plan that looks best on average, the authors use a “robust” rule that tries to avoid big regrets — the large losses you would see if a chosen policy turned out to be the wrong one. Applying that rule to a wide set of climate‑economy models favors faster reductions in carbon dioxide emissions than a simple average would.

The team built 100 different integrated assessment model (IAM) structures. Each IAM is made of four parts: a socioeconomic module (things like population, economic growth, and emissions), a climate module, a model of the costs to cut emissions (abatement costs), and a damage function that relates temperature to economic harm. They sampled uncertain parameters inside each model with Monte Carlo runs, found the best mitigation pathway for each model by maximizing expected welfare, and then compared all 100 candidate policies using a minimizing maximum regret (minimax regret) criterion. That means they ranked policies by how large the worst regret could be across many possible futures.

Using this approach, the most robust policy in their sample calls for full decarbonization by 2050 and limits the expected global mean temperature anomaly to about 2.2°C above pre‑industrial levels across the considered futures. The key reason for this result is an asymmetry in the consequences of doing too little too late: some combinations of high emissions pathways and damage relationships produce much worse outcomes if mitigation is weak. Being regret‑averse therefore pushes toward precaution and earlier, steeper cuts in emissions.

This matters because many policy analyses rely on single models or on simple averages across models, which can hide the risk of very bad outcomes. The robust decision‑making framework keeps the different model results separate, so decision makers can see how strategies perform across a wide range of plausible futures and which uncertainties matter most for the choice. That can help prioritize safeguards against the outcomes that would produce the largest regrets.