Pessimistic inductions are a family of arguments that run as follows. First, they identify some epistemologically contentious domain. Second, they identify an epistemically negative property that past attempts to theorize about the domain would have if each of them were compared to our contemporary efforts. Third, they cite this historical pattern as a track record that bears inductively on our contemporary efforts to construct and identify accurate theories. Fourth, they perform an inductive inference to conclude that our contemporary situation probably has that negative property too. Finally, they conclude that we should retract our confidence in contemporary theories due to this looming background statistic. This conclusion is supposed to follow despite the fact that our current first-order evidence favors our current theories. After all, every predecessor in the track record was in its day supported by the best evidence then available, and yet has since been discarded. Pessimistic inductions emphasize a simple point: any pronouncement about the epistemic status of current theories must take their own track records into account.
Chapter 1 discusses not only the nature of pessimistic inductions but also their ubiquity. While they remain some of the most widely cited challenges to scientific realism since the 1970s, their pedigree goes as far back as the 1500s. They have been used by Montaigne, Locke, Hume, Pareto, and even appear in Jonathan Swift’s classic tale, Gulliver’s Travels. I also discuss their wide applicability. They have been used against descriptive theories of reference, both physicalist and dualist theories of mind, the legitimacy of folk concepts, and even metaphysics more generally. One needn’t be a philosopher of science to have a stake in the success or failure of pessimistic inductions.
Chapter 2 introduces what is now known as the Classical Pessimistic Induction. It was made famous in the late 1970s by Hilary Putnam and still has its defenders to this day. Its structure: past empirically successful theories have typically been radically false by current lights; current theories aren’t relevantly different; so, they’re probably radically false too. I argue that this prominent argument is defeated by the following three principles of evidence:
· If our hypothesis predicts the data, then that data can’t disconfirm our hypothesis.
· Some correlations are merely apparent in light of new underlying information.
· Inductions are legitimate only if their samples are random.
I show that the Classical Pessimistic Induction runs afoul of these three principles by virtue of its own inferential structure. The problem stems from its reliance on a method of historical comparison with contemporary theory to interpret the truth values of historical theories in the track record.
Chapter 3 discusses an influential argument by Kyle Stanford called the New Pessimistic Induction. It runs as follows. Past scientists have typically been unimaginative—they routinely fail to conceive of serious theoretical alternatives that are at least as well supported by the phenomena as the ones that scientists actually endorse; current scientists aren’t relevantly different; so, they’ve probably failed to imagine serious alternatives too. Advocates of Stanford’s argument contend that it escapes the standard roster of objections raised against the Classical Pessimistic Induction. I argue that Stanford’s argument succumbs to the same types of problems that I highlight in the previous chapter. After all, it too requires a method of historical comparison with historical successors to interpret its track records. So, the New Pessimistic Induction likewise runs into problems with the principles of evidence above.
Chapter 4 replies to the most recent pessimistic induction by Gregory Frost-Arnold—the Misleading Evidence Induction. It argues that past evidence has typically been misleading, that current evidence isn’t relevantly different, and that therefore our current evidence is probably misleading too. I survey a variety of available analyses of misleading evidence and argue that no matter how we construe the Misleading Evidence Induction, each interpretation violates at least one of the principles of evidence above. Once again, the argument relies on the very same method of historical comparison that bedeviled the pessimistic inductions from the previous chapters.
I conclude that pessimistic inductions against scientific realism fail a priori. My treatment of this issue is surprisingly novel. It is novel, in that none of the authors who have written on the topic since the 1500s have developed this solution. It is also surprising, because I only use simple principles of evidence accepted by all parties to the dispute in conjunction with features intrinsic to pessimistic inductions to resolve the issue.