Climate engineering measures are designed to either reduce atmospheric carbon concentration (by growing trees or spreading iron in the ocean, for example) or directly influence the radiation reaching or leaving the earth (by injecting sulfur into the stratosphere or modifying cloud formations, for example) to compensate for greenhouse gas–induced warming. The former measures are termed carbon dioxide removal (CDR), which we characterize as a low-leverage causative approach, and the latter are termed radiation management (RM), which we characterize as a high-leverage symptomatic approach. There are similarities between CDR and emission control. Accordingly, benefit-cost analysis can be used to assess certain CDR measures. By contrast, high-leverage RM represents a genuinely new option in the climate change response portfolio, at first glance promising insurance against fat-tail climate change risks. However, the persistent intrinsic uncertainties of RM suggest that any cautious climate risk management approach should consider RM as a complement to (rather than a substitute for) emission control at best. Moreover, the complexity of the earth system imposes major limitations on the ability of research to reduce these uncertainties. Thus we argue that a research strategy is needed that focuses on increasing our basic understanding of the earth system and conducting comprehensive assessments of the risk(s) associated with both climate change and the deployment of climate engineering.