Information and Experimental Knowledge

By James Mattingly

An ambitious new model of experimentation that will reorient our understanding of the key features of experimental practice.

What is experimental knowledge, and how do we get it? While there is general agreement that experiment is a crucial source of scientific knowledge, how experiment generates that knowledge is far more contentious. In this book, philosopher of science James Mattingly explains how experiments function. Specifically, he discusses what it is about experimental practice that transforms observations of what may be very localized, particular, isolated systems into what may be global, general, integrated empirical knowledge. Mattingly argues that the purpose of experimentation is the same as the purpose of any other knowledge-generating enterprise—to change the state of information of the knower. This trivial-seeming point has a non-trivial consequence: to understand a knowledge-generating enterprise, we should follow the flow of information. Therefore, the account of experimental knowledge Mattingly provides is based on understanding how information flows in experiments: what facilitates that flow, what hinders it, and what characteristics allow it to flow from system to system, into the heads of researchers, and finally into our store of scientific knowledge.

• Language: English
• Publisher: University of Chicago Press
• Release date: Dec 13, 2021
• ISBN: 9780226804781

Book preview

PART I

Aspects of Experimentation

1

Introduction to Part I

1.1. About Induction

I cast some aspersions above on the idea that knowledge has anything interesting to do with justification. I will have a lot more to say about that in part II when I rehearse and defend Dretske’s basic story of knowledge as information-caused (or -sustained) belief. Now, though, I want to clarify why justification cannot be a legitimate aim in experimentation, or in scientific inquiry generally. I believe that an important source of our preoccupation with justification when thinking about experimental knowledge is Boyle and his use of the matter of fact as the foundation of the experimental method.¹ For Boyle, the matter of fact was the cornerstone on which all further experimental claims would be built. To work as a foundation of that sort, the matter of fact would have to be beyond question and would thus need to be generated through a process that, when done properly, could not fail. Of course there is no such process. Any activity, even when done properly, can sometimes fail, but the rhetoric supporting the method had a heavy burden to bear.²

This rhetoric is needed to get the experimental method up and running as a sustainable enterprise, and it contains a very strong commitment to assurance, security, and stability—in short, to justification as a constitutive feature of our experimental knowledge. What goes wrong then—wrong from the perspective of our attempts to understand experiment, but not from the perspective of what was needed to get the process of modern science up and running—in the transition beginning in the Renaissance to the philosophy of nature that leads to the Royal Society and from there to our current conception of the ultimate foundation of scientific knowledge, is the conflation of experiment itself with a motley panoply of justificatory techniques. It is understandable that when one of your main tasks with respect to your novel experience-based conception of natural philosophy is to promote and to defend that conception, the result will be strongly oriented toward an epistemology founded on assurance. An important part of the business of the mechanical philosophers was to convince the consumers (and producers) of the new natural philosophy that the method could be trusted to give real, secure knowledge of nature. So it is no wonder that security of results was built into the very conception of what it was to be doing things correctly. It seems that even now the main currents of philosophy of science equate experiment with the epistemological force of experiment, or at least make a very fine distinction between them. Of course this view is not explicitly articulated very often, and by no means is it universally held. However, given the lack of attention to experiment generally in the philosophy of science during the period from, say, the 1920s until the 1990s, there has also been little explicit probing of this implicit view.

We no longer think that any of our experimental claims are absolutely certain, nor do we think that such a lack of certainty is any threat to our scientific knowledge. Indeed, science is often understood to be essentially a self-correcting enterprise that upholds the revisability of all claims as a central tenet. Yet we still tend to think of the category of scientific fact as containing only truths on the one hand and constituted by something like Boyle’s method on the other. And we do not notice that Boyle’s category of fact covers all claims made in accordance with the demand of proper performance of a trial witnessed by disinterested audiences of good character.³ In addition to that muddle, much of the contemporary literature on the foundations of scientific experiment is concerned with the security of experimental claims and other justificatory notions. These notions, some of which are mentioned above, include: assurance, security, stability, trustworthiness, etc. In short, they are notions that indicate not whether the claims being made are true, nor whether the beliefs they report are properly acquired, but rather what resources a knower has available to defend these claims and to persuade others of their truth. But in my view the latter (justificatory features) need to be kept clearly separate from the former (broadly informational features). I will in due course defend an information theoretic account of knowledge that makes no appeal at all to the things a knower needs to possess or be able to do in order to defend her claims. Naturally there are some who continue to use the language of justification even in the context of externalist accounts of knowledge. However, in my view that simply leads to confusion. Consider: justification, at least on traditional accounts of knowledge, leads to trouble in part because any justification invites the further question of its own justification, and so on. This leads to the well-known trilemma posed by knowledge skeptics: infinite regress, dogmatism, circular reasoning. Part of the motivation (as we will see later) for information as the fundamental principle is as replacement for the traditional notion of justification. One of its key virtues is that it does not lead to the trilemma because it is not about persuasion, or defense, or even knowing that one knows. An account of this sort that generates knowledge without justification is naturally immune to skeptical worries, as I argue elsewhere and make clear in the remainder of this book.

Eschewing justification and its concomitant notions, my account is at odds with much of what one sees in epistemological accounts of scientific experimentation. For when philosophers and foundations-oriented scientists speak about the knowledge claims resulting from experiments, they are attempting to provide analyses of how various standards of evidence can be brought to bear to evaluate those knowledge claims, and they are not principally concerned with the nature of experimental knowledge as such. I am going to argue in this chapter that much of the philosophy of science involving experimental knowledge over the last few decades is directed toward an analysis of various classes of justification and not toward an answer to the question of what conceptually is going on when we generate experimental knowledge.

My preoccupation in the bulk of this book is with exactly that question, and so I mostly do not engage directly with justificatory questions. That said, this chapter is an attempt to situate the concerns of the book against the broader background of recent philosophy of science, especially as its concerns relate to experiment. There are several important issues in the philosophy of science in the background of my treatment here, and my approach to these strongly colors my approach to experiment more generally. Perhaps surprisingly, my view of theories (as a hybrid of their formulations and the mathematical models they pick out) provides important structure, as does my view of the nature of scientific explanation. For the latter, I have adopted what I take to be the best reading of Hume’s version of scientific explanation—that explanations are cognitively salient for us only when they are couched in terms of local causal interactions. On the other hand, I have no view at all on the perennial realism/antirealism debate in the philosophy of science, other than that it’s a red herring distracting us from our real quarry, an account of scientific knowledge. Issues in the confirmation of scientific theories, and the role of experiment in that confirmation process, are downplayed here, and would have to be recast in order to fit them properly into my own account of experimental (and more broadly, scientific) knowledge. I will begin, though, with a polemical discussion of the contexts of justification and of discovery.

1.2. Contexts of Research

The first shocking claim I wish to make is that there is no scientific context of justification for empirical knowledge generation. The second is that even if there were, all experimentation would operate not in that context but instead in the context of discovery. My reason for the first claim is simply that there is no such thing as empirical justification at all, and particularly not in the epistemology of the sciences. It was and is a mistake to treat the business of science as securing rather than producing knowledge. The former is properly the domain of policy, while the latter is the real aim of the sciences. My reason for the second claim is is elaborated in detail in the rest of this book, but in short, some experiments are exploratory, obviously for discovery, and the rest are for letting information about the world flow through them to scientific researchers. There is simply no place for justification in this enterprise.

The claim that there is no empirical justification shouldn’t be shocking, but it is. So a good place to start my discussion of contemporary understandings of experiment might be with Karl Popper, who told us that we could not have any knowledge of nature that went beyond bare token instances.⁴ His reason for this outlandish claim is that there are basically two kinds of things we could conceivably know about nature: (1) that something has some property, and (2) that generally things of one type are also of some other type. More concretely, I can know that the object in front of me is rigid, blue mixed with clear, sphondiloidal at one end, etc. I could conceivably also know that dogs are mammals, domesticated, carnivorous, etc. That’s about it. But the latter relies on inductive logic, and there can be no justification for that. If what we mean by knowledge is something involving justification, then we have very little of it.

The falsificationist fiasco does not come from a mistake about justification, and its impossibility when induction is involved. Popper is quite right in his view that no inductive claim can ever have enough evidence in its favor to be beyond doubt—for if it did it would be deductive. He is also right that Hume’s analysis of induction generally, the necessary step in any inductive argument, shows without question that there can be no evidence for induction that doesn’t itself appeal to induction. Hume drives this point home, and yet it is still often underappreciated. Induction can never be justified—at all. There has never been and will never be a single shred of evidence in favor of induction. We may, and do, succeed with its use; we may well even be correct in the conclusions we draw when we use it; but its justification would need to rely on inductive reasoning itself, and indeed on a justified use of that reasoning. Where would that come from? Only induction can provide justification for induction, and it’s the kind of thing that, if you need it, you can’t have it. Here Popper is on solid ground. His real mistake, and the one we continue to make even after rejecting his model of scientific development, is to think that justification pertains to knowledge, rather than to persuading others of our views or to legitimizing the publication of those views. In laying out the problem of induction, he tells us that it can also be formulated like this ([1959] 2005, 9):

For many people believe that the truth of these universal statements is known by experience; yet it is clear that an account of an experience—of an observation or the result of an experiment—can in the first place be only a singular statement and not a universal one. Accordingly, people who say of a universal statement that we know its truth from experience usually mean that the truth of this universal statement can somehow be reduced to the truth of singular ones, and that these singular ones are known by experience to be true; which amounts to saying that the universal statement is based on inductive inference.

Popper and others thought that without justification there could be no knowledge, and those others’ disagreement with Popper about whether we really do have scientific knowledge set the stage for a fruitless attempt to provide justifications for inductive claims. But as we will see below, there is available a much more congenial account of scientific knowledge, one that does not appeal to justification. In contemporary science, we are often making universal claims about nature—claims that support counterfactuals across all of space and time. There is always the possibility that our conclusions are wrong. We keep looking, and attempting to get more data, and higher-order data as well (that is, data about the nature of our original data). But it is not right to say that we do not know because the data are not all in. Rather, we may well know, and indeed we may well know based on a very small amount of data. In point of fact, any amount of data we collect will be a small amount, since generally we’re making claims with infinitely many possible token instances, and all finite quantities are small compared to infinity. In addition to the circularity problem above, there is then an entirely separate issue with the so-called inductive sciences that comes down to just this point. If the game is to get enough data to justify our claims, then we are out of luck, for from that perspective our data are finite and thus have no significant justificatory weight. For we are making universal claims, claims that in principle apply to an infinity of situations, and all finite quantities are exactly as good as zero when stacked up against infinity.

Yet we do know things about nature. So if we wish to analyze empirical knowledge, knowledge which is acquired only inductively in the end, we must immediately reject any account of knowledge founded on justification. Accepting that we do have empirical knowledge, and accepting the trivial logical point above, entails that we cannot accept any model of scientific knowledge of which justification is a necessary component. We must, when it comes time to share the results of our scientific inquiries, make reference to standards of acceptance, to the strength of the data, and so forth. But those are only, and can only be, pragmatic norms of acceptance. They can never be part of what we mean when we say that we have scientific knowledge.

Rejecting the idea that scientific knowledge has any connection to justification dispenses with other concerns that have troubled philosophers. For example, one issue that has dogged the foundations of science for a long time is whether there can be crucial experiments in the sciences, and more generally how we are to respond to the problem posed by Duhem and revived by Quine: that I can maintain any self-consistent hypothesis I wish no matter what data I find that are supposed to bear on that hypothesis, provided I am willing to adopt sufficient auxiliary hypotheses. However, the worry about crucial experiments—crucial in the sense of decisive—makes sense only against a justificationist backdrop. To be sure, we do want guidance on which directions of inquiry are most likely to be fruitful going forward. So it might seem that we could use experiment to settle definitively the question of which theory is correct (if either). We can! But not because the data stop me from holding on to my theory, come what may. The data don’t stop me. Instead my commitment to understanding what information about nature those data are really carrying stops me. Strategies of justification fail in the face of Duhem–Quine problems, but the right thing to do is to stop thinking of knowledge claims as having to do with justification. Philosophers are accustomed to thinking that knowledge and justification go together. They don’t, they can’t, and happily they don’t need to.

I’m going to begin the book proper with what I take to be a few obvious points. Then I’m going to use those obvious points to motivate a view of experimental knowledge that also seems reasonably obvious, or at least (once it is made clear) should seem to be a plausible account of what we already think is going on in experiments. But then I will use that account to do two things: (1) to undermine some standard stories about the nature of experimentation and how it connects to our knowledge of nature, replacing those stories with one about the standards of communication of knowledge; and (2) to argue that much more counts as experimental knowledge than is commonly allowed, and to show that that fact is an easy consequence of my view.

The most obvious point I want to make is that we do have experimental knowledge. That is to say, some of the things that we do really count as knowledge-generating experimental investigations. Thus, perhaps, we know that dioxins cause liver cancer in rats; that saccharin causes bladder cancer in humans; that cathode rays are streams of charged particles; that certain arrangements of very cold particles support sound waves obeying the Gross–Pitaewski equation, thus giving knowledge about the cosmos at large; that agents in perfect knowledge game situations tend to maximize their utilities; that similar agents tend to value objects more when they own them than when they don’t; etc. On the other hand, we might well not know some of these things: it might be that some of those claims are false, while others are true; it might be that some are true but we still do not know them. In any case, though, I do think that at least sometimes we know things on the basis of our experimental practices. I also think I might know how to argue with someone—a skeptic, for example—who thinks that we do not, but I am not going to try. That’s a different project.

Even though I won’t give a concerted argument against skepticism, I will say something on this point. Let’s return to induction, because induction is at the heart of all scientific knowledge. Does it work? Well, it has worked in the past, and I believe it will continue to work in the future. Probably I am using induction to generate that belief, and if induction does work then that’s OK. Probably we all believe that induction works. Even if, like the skeptics, we are not willing to affirm explicitly that it works, still every action we take that is guided by our past experience betrays an implicit trust in induction: bread nourishes, it doesn’t poison, so we eat it; things fall down, not up, so we avoid walking underneath construction cranes; mustard is delicious and ketchup is disgusting, so we put the one on hotdogs and the other in the trash⁵; etc. In addition to our individual beliefs that induction works, let’s suppose that it does indeed work. But even assuming that we believe it, and that it works, do we really know that it