Investments in climate science come with an expectation of social benefit. Science policy--decision processes through which individuals and organizations support, manage, and evaluate research--plays an important role in determining those outcomes. Yet the details of how climate science policy actually works have received very little attention amid academic and policy-focused discussions of climate science. This dissertation examines climate science policy with particular attention to how it supports "public values" that justify research investments. It is widely recognized funding for climate science in the US has advanced knowledge considerably in recent decades but failed to produce useful information for decision makers. In Chapter 2, I use a methodological approach known as Public Value Mapping (PVM) to investigate this failure of the science policy system. My results show that science funding institutions have been ineffective at guiding climate science toward desired outcomes because of problematic, but common assumptions about the links between science and societal benefit. The remaining chapters look more closely at the implications of these tacit assumptions, which are held by individuals, and embedded in the organizations that implement climate science policy. Chapter 3 examines the notion that prediction is essential to climate science. Wide acceptance of the "prediction imperative" limits the scope of climate science policy. Chapter 4 examines the interplay of values and assumptions in two recently established organizations in Australia, each supporting research on climate change adaptation. In Chapter 5 I document a widespread assumption in the climate science literature that agreement among multiple models should bolster confidence in their results. This can only be correct if the models are independent of one another. Climate scientists have not demonstrated this to be true, nor have they offered a plausible framework for doing so. This dissertation adds an important dimension to our understanding of how climate science knowledge is produced, while offering constructive and practical recommendations to science policy decision makers working in government programs that fund climate science. Insight from these chapters suggests that an explicit and reflexive focus on values in science policy can be helpful to organizations pursuing science policy innovation.