Values and uncertainties in climate prediction, revisited

Abstract

Philosophers continue to debate both the actual and the ideal roles of values in science. Recently, Eric Winsberg has offered a novel, model-based challenge to those who argue that the internal workings of science can and should be kept free from the influence of social values. He contends that model-based assignments of probability to hypotheses about future climate change are unavoidably influenced by social values. I raise two objections to Winsberg’s argument, neither of which can wholly undermine its conclusion but each of which suggests that his argument exaggerates the influence of social values on estimates of uncertainty in climate prediction. I then show how a more traditional challenge to the value-free ideal seems tailor-made for the climate context.

Introduction

In a classic paper, Richard Rudner (1953) argued that ethical values are a required part of the internal workings of science. Since evidence never establishes a hypothesis H with certainty, scientists must decide whether the evidence is sufficiently strong to merit acceptance of H, and this decision reflects judgments about how bad it would be (in a typically ethical sense) to be mistaken in accepting or rejecting H. Thus, for example, we would demand stronger evidence before accepting the hypothesis that a new medical procedure for children is safe than before accepting the hypothesis that a manufactured lot of shampoo bottles meets desired quality specifications.

In a classic reply, Richard Jeffrey (1956) challenged Rudner’s conclusion by denying that scientists are in the business of accepting or rejecting hypotheses in the first place. According to Jeffrey, scientists’ job is to assign probabilities (degrees of belief) to hypotheses; it is then up to decision makers—with their social, political and ethical value commitments—to take action or not in light of those probabilities, that is, to determine whether the probabilities provided by scientists are high enough to warrant various courses of action. On what we might call a Jeffreyan view, the internal aspects of scientific research—including the assignment of probabilities to hypotheses—can and should remain free from the influence of social values.¹

Today, philosophers of science remain divided on the appropriate roles of ethical and social values in science. Some, like Douglas, 2000, Douglas, 2009, argue that such values are a required part of the internal workings of science in much the way that Rudner suggested. Others, like Sandra Mitchell (2004), seem to embrace something much closer to the Jeffreyan view. Recently, Winsberg, 2010, Winsberg, 2012; see also Biddle and Winsberg, 2009) has offered a novel challenge to the Jeffreyan view. He argues that, when it comes to hypotheses about future climate change, it is infeasible for climate scientists to exclude the influence of social values when assigning probabilities. As he puts it: “Scientists cannot assign probabilities to hypotheses about future climate change—or, more specifically, estimate the uncertainties of climate predictions—in a manner that is free from non-epistemic considerations …” (2010, p. 119).²

To begin to see what Winsberg has in mind, consider the following statement: The probability is 0.90 that global temperature will rise by between 2 and 3 °C during the 21st century if greenhouse gas emissions continue at current rates. This statement assigns a probability to a hypothesis about future warming. It is also a way of expressing uncertainty about how much warming would occur; instead of reporting that the temperature would definitely rise by, say, 2.1 °C, an assignment of probability is made to an interval of change. Winsberg argues that social values unavoidably influence probabilistic uncertainty estimates like these. What is especially novel about his argument is its grounding in the practice of complex simulation modeling: assignments of probability to hypotheses about future climate change are influenced by social values, he argues, because of the way these values come into play in the building and evaluating of climate models, on whose results the assignments of probability depend.

The present paper has two main aims: first, to critically evaluate Winsberg’s model-based challenge to the Jeffreyan view and, second, to show how a more traditional set of replies to Jeffrey seems tailor-made for the climate context. Section 2 provides background on climate modelling and on uncertainty in climate prediction. Section 3 presents a distilled version of Winsberg’s argument. Section 4 raises two objections to the argument; while neither can wholly undermine his conclusion that social values have some unavoidable influence on uncertainty estimates in climate prediction, each suggests that social values exert less of an influence than he implies. Section 5 revisits a more traditional set of replies to Jeffrey. Without necessarily endorsing these replies, it is shown that they dovetail nicely with recent discussions—internal to climate science—about the estimation of uncertainty in climate prediction.