Robust decision making (RDM) is a framework for making decisions with a large number of highly imperfect forecasts of the future. Rather than relying on improved point forecasts or probabilistic predictions, RDM embraces many plausible futures, then helps analysts and decision makers identify near-term actions that are robust across a very wide range of futures--that is, actions that promise to do a reasonable job of achieving the decision makers' goals compared to the alternative options, no matter what future comes to pass. Rather than asking what the future will bring, this methodology focuses on what we can do today to better shape the future to our liking.
RDM emerged from work at RAND beginning in the early 1990s, when we, analysts Robert Lempert and Steven Popper, were separately grappling with policy problems characterized by deep uncertainty and potentially non-equilibrium dynamics--in particular, climate change and the transition of east European communist societies to market economies. Meanwhile, RAND computer scientist Steve Bankes was grappling with the question of how one can use imperfect computer models to inform policy decisions, particularly to deal with the next wars rather than previous ones.
In brief, RDM uses the computer to support an iterative process in which humans propose strategies as potentially robust across a wide range of futures. The computer then challenges these strategies (stress tests) using simulations and data extrapolations to suggest futures where these strategies may perform poorly. The alternatives can then be revised to hedge against these stressing futures, and the process is repeated for the new strategies.
Rather than first predicting the future in order to act upon it, decision makers may now gain a systematic understanding of their best near-term options for shaping a long-term future while fully considering many plausible futures. The result is near-term policy options that are robust--i.e., that, compared to the alternatives choices, perform reasonably well across a wide range of those futures.
The strength of robust decision making is its flexibility. In this iterative process, the computer retains the full range of uncertainties, multiple interpretations, and other ambiguities and can bring key bits of information to decision makers' attention at any point where it might help distinguish among the merits of alternative decision options. This process can help break down institutional barriers to considering multiple futures, because it provides systematic criteria for determining which futures ought to be considered. It can help decision makers avoid "over-arguing," which occurs when decision makers pretend they are more certain than they actually are to avoid losing credibility in policy debates--by allowing them to acknowledge multiple plausible futures and to make strong arguments about the best policies for hedging against a wide range of contingencies.
Computer-supported RDM at its root combines the best capabilities of humans and machines. Humans have unparalleled ability to recognize potential patterns, draw inferences, formulate new hypotheses, and intuit potential solutions to seemingly intractable problems. Humans also possess various sources of knowledge--tacit, qualitative, experiential, and pragmatic--that are not easily represented in traditional quantitative formalisms. Working without computers, humans can often successfully reason their way through problems of deep uncertainty, provided that their intuition about the system in question works tolerably well.
Using their talent for storytelling, humans can challenge each other with "what if" scenarios to probe for weaknesses in proposed plans. These processes succeed because the best response to deep uncertainty is often a strategy that, rather than being optimized for a particular predicted future, is well hedged against a variety of different futures and evolves over time as new information becomes available. …