Scientists have turned their attentions to computer modelling as a method for studying global systems in order to evaluate claims of potential catastrophe and to see whether intervention is possible. In chapter 2, I will discuss the nature of modelling in general. In this chapter, I will sketch some well-known models of global systems designed in response to fears of global catastrophe and supply some of their history.

The models I am interested in each deal with a unique problem. These models of global systems examine the potential of some human activities to result in catastrophe. Nuclear winter, ozone depletion,...

In the previous chapter, we acquainted ourselves with four contentious models of global processes. In the next chapter, we will scrutinize the same four models to see whether any of them generate predictions reliable enough to support decisions about economic or political policies. In this chapter, we will try to understand the nature of models and how models are used in science. What is a scientific model, and what do we do with one?

The main function of scientific models is to provide predictions and what Mary Hesse calls “intellectually satisfying” explanations.¹ A scientific model is a representation of the...

It is clear from the examples given in chapter I that global models are porous with uncertainty. When, for instance, we consider fluctuations in solar radiance, complicated causal chains of feedback, and other variables, we find that global climate models are just as likely to under estimate global warming as overestimate it – and, therefore, might not be much good at all. How, then, are we to assess the predictions of catastrophe stemming from these models? Can we know (or have good reason for believing) that a model of a global system, from which predictions of the future are generated, is...

The three conditions oulined in chapter 3 are meant to serve as criteria for assessing whether a prediction is a good test of a model, theory, or hypothesis. I have pointed out already that for models of global systems that are predicting catastrophe, this direct testing procedure is not useful. We should like to have some method of assessing whether a model is capable of making reliable predictions without having to wait and see if those predictions come to pass.

We can start on this task by assessing a model’s ability to satisfy two requirements of successful prediction. In order...

I have pointed out that standard ways of evaluating scientific models by trial and error do not meet the requirements for distinguishing accurate global models from the not so accurate ones when predictions of catastrophe are involved – because we need to make the distinction without having to see whether predictions of catastrophe come to pass. Given that models of global systems cannot be tested by direct methods of prediction and trial and error, and that we are not sure whether some particular model satisfies the two requirements for successful prediction, how are we to defend predictions based on models of...

The examples in chapter 5 show that indirect tests of global models should not generate confidence in their predictions. These arguments were not addressed merely to the general epistemological problems that attend all belief. All beliefs are uncertain in a general, skeptical sense.The activity of projecting trends into the future, for example, involves deep Humean problems about induction. However, if my criticisms of models of the future were all of this strong and far-reaching kind, they would be too recondite to be relevant to the particular issues I am discussing. My concern is not with skepticism and anti-skepticism, but with...

The problem raised by models of global systems that predict catastrophe involves two questions: what to believe and what to do. The first is epistemic: what should we believe? The first part of this book was concerned with what good reasons there are for accepting predictions supplied by models of global systems. Our beliefs about the accuracy of these predictions will remain uncertain for some time. We are uncertain, therefore, whether to take seriously recommendations for action based on model representations, even if such action might prevent out comes we would consider catastrophic. The second question is practical: what, then,...

When probabilities are very uncertain or entirely unknown, the preferred strategy among decision theorists is to avoid deciding what to do until more information has been gathered and the uncertainty has been reduced. However, as I said in the previous chapter, were we to adopt this strategy in cases of global risk involving potentially irreversible catastrophes, the very purpose of making a decision would be defeated. In such cases we are forced to make a decision one way or the other because we cannot afford not to decide, and we cannot afford to wait for more evidence because, as we...

We saw in the previous chapter that, on the one hand, the precautionary principle has seemingly intractable problems and that, on the other, epistemic and practical conservatism might lead to global catastrophe because it encourages inaction in the face of enormous risk. In chapter I0, I will come back to the problems with the precautionary principle and offer a defence of it. First, though, I want to discuss another way of responding to the lack of definite probabilities for predictions of global risk. Some authors argue that, if normal accounts of rationality do not give us reasons to act precautiously,...

The basic thrust of arguments in favour of acting precautiously is that there is too much at stake for us to do otherwise. The disposition to err on the side of caution seems so commonsensical to so many people, that it is often thought not to be in need of justification. When there is a possibility of catastrophe, it is best to eliminate the chances of that possibility. However, this often-stated argument is, perhaps, too vague and simplistic–for reasons outlined in chapter 8. It is rejected by many as unscientific for its reversal of the burden of proof. The...