It is increasingly recognized that decisions regarding climate change policy cannot be exclusively framed in technical terms; interdisciplinary approaches, incorporating social, economic and ethical dimensions, are increasingly deemed necessary for good policymaking. While many problems can/are being framed in an appropriately broad context, some significant policy decisions still primarily focus on technical predictions of climate change that are made by general circulation models (GCMs).

To provide a tool to further motivate interdisciplinary approaches to climate policy as well as to frame uncertainty in modeling for instances when technical approaches are necessary, I will develop the philosophical framework of robustness as an epistemic strategy for understanding complex systems. This framework can help to evaluate the effectiveness of existing climate science research and suggests directions for future interdisciplinary work that proceeds from a solid understanding of independence and the possible limits of models.

Biologist Richard Levins first developed the robustness framework in his 1966 “The Strategy of Model Building in Population Biology.” After recognizing that “all models leave out a lot and are in that sense false, incomplete, inadequate,” Levins advocates the development of independent models as a means by which to analyze complex systems. To illustrate the development of robust conclusions, Levins explored how multiple approaches from conceptually independent sub-disciplines were combined in order to establish the validity of the overall field of population biology. Levins says the resulting "truth is the intersection of independent lies," and advocates that examining complex systems through independent, idealized models is a powerful way to explore complex phenomenan that are analytically intractable by other means. Given that GCMs are perhaps the most prominent strategy of attempting to understand change throughout the globe, application and development of the robustness strategy used by Levins is directly relevant and indicates research directions for the climate science community.

For example, some philosophers have raised questions about the independence of GCMs (Wendy Parker, 2004), and a robustness framework encourages potentially fruitful research on independence between climate models. Given that the agreement of multiple GCMs is often taken as evidence that an event will occur, to what extent are the conclusions of climate science robust? Are the models so highly interconnected, in code, basic equations, and fundamental modeling strategies, such that they would not be considered independent in Levins' sense? A robustness framework would provide a lens for analysis of the independence between different climate models and would second calls (such as those made by social scientist Simon Shackley) for the development of multiple, independent means for understanding climate and for framing climate policy decisions.

Further, a robustness framework also brings forward questions about the limitations of models, especially as regards long-term predictions such as those made by GCMs. Levins argues that the modeling of complex systems always introduces tradeoffs between generality, precision, and realism. Following from a discussion about mathematical complexity, Levins argued that within complex systems a model cannot possibly produce predictions that are simultaneously general, precise and accurate. Models can only attain two of those variables, such as maximum realism and generality, and will be lacking in the third variable, here precision. There is a vigorous debate about Levins' tradeoff hypothesis, and whether modern computing power could allow for models that attain all three desiderata. Everyone recognizes that GCMs today are highly uncertain, but some intend to use GCMs to create accurate long term predictions that are global in scope and which provide information about precise (local) climate changes. If Levins' tradeoff hypothesis is correct, then the utilization of GCMs for such ambitious goals will be a failure. Given that there is good reason to believe that Levins' hypothesis will hold true of GCMs for the foreseeable future, scientists and policymakers should look more favorably upon predictions that rely on a strategy of recognizing the tradeoffs Levins describes. Most importantly, however, consideration of the tradeoff hypothesis further underscores that climate policymaking must rely upon considerations that extend beyond the technical details of individual models.