Theory and Reality
Page 33
The third follows from the last section of chapter i z. I used a broad concept of "representation" to describe the relation that science aims to achieve between theories and reality. I resisted the tendency to make concepts from the philosophy of language, like reference and truth, central in this part of the story. I also emphasized the role of models. But we do not have a good philosophical theory of representation yet, and even the most basic issues here are fraught with controversy. A big cloud of uncertainty still hangs over this part of philosophy of science.
Along with these clouds of uncertainty, however, I think we can see some fairly definite progress in the philosophy of science in recent years. The idea that, in some way or other, all science is concerned with is the description of patterns in experience has finally been (mostly) abandoned. Scientific realism has been developed and defended in sophisticated forms. The field has become less dominated by questions about language, and proper attention is being paid to modelbuilding as a crucial part of scientific work. Theories of testing and evidence are in vastly better shape than they were fifty years ago. The very idea of looking closely at the relation between the reward structure in science and epistemological issues is a crucial advance. So there has been progress, but there is still much to do.
For definitions and quick discussions of other philosophical terms, I recommend Simon Blackburn's excellent book The Oxford Dictionary of Philosophy. Blackburn also gives more detail on many of the terms discussed here.
Sometimes confusion in newcomers to philosophy arises not from technicality, but from slightly different philosophical uses of everyday language. For example, in philosophy the word "strong" when applied to a view or a hypothesis does not mean effective, and it carries no positive (or negative) connotation. "Strong" means something more like extreme, bold, or tendentious. This is related to the use of the term in logic, where a strong claim is one that has a lot of implications. "Weak" in this sense means cautious, hedged, or moderate. Scientists sometimes uses "strong" in the same way. So a "strong" version of a view (empiricism, realism, etc.) is not necessarily better than a weak form. Confusingly, philosophers do sometimes say strong argument when they mean that the argument is good, or convincing.
After discussing each term below, I indicate the chapters or sections of the book in which the term is important. Terms in boldface have their own entry in the glossary.
There are two terms used in this book that are my own modifications of more standard terms. These are "explanatory inference" and "eliminative inference." In both cases I am avoiding overly broad use of the term "induction."
Abduction. One of the many terms for explanatory inference. This one was coined by C. S. Peirce. (3.2, 14.5)
Analytic/Synthetic Distinction. Analytic sentences are true or false simply in virtue of the meanings of the terms within them. Synthetic sentences are true or false in virtue of both the meanings of the words and the way the world is. Logical positivism treated this distinction as very important. Quine argued that it does not exist. (2.3, 2.4, 2.5)
Anomaly. In Kuhn's theory of science, an anomaly is a puzzle that resists solution by the methods of normal science. This is close to the word's ordinary meaning (roughly, something out of place). (5.4)
A Priori/A Posteriori Distinction. If something is known (or knowable) a priori, it is known (or knowable) independently of evidence gained via experience. Knowledge that relies on evidence from experience is a posteriori knowledge. (2.3)
Bayesianism. The theory of evidence and testing that gives a central role to Bayes's theorem, which is a provable result in probability theory. Bayesians treat all rational belief change as a matter of updating one's degrees of belief in propositions, in accordance with the principles of probability theory. (chapter 14)
Confirmation. A relationship of support between a body of evidence and a hypothesis or theory. Confirmation is not the same as proof; a theory can be highly confirmed and yet be false. Logical positivism and logical empiricism put much emphasis on the role of this relationship in science, usually trying to analyze it with an "inductive logic." Their attempts were not very successful. (chapters 3, 4, 14)
Constructivism (Social Constructivism, Metaphysical Constructivism). A word with many meanings. In the debates discussed in this book, "constructivism" usually refers to some sort of view in which knowledge (and sometimes, reality itself) is seen as actively created by human choices and social negotiation.
People advocating constructivist views often do not distinguish carefully between the view that theories (or classifications, or frameworks) are constructed, and the view that the reality described by those theories is constructed. I use the term "metaphysical constructivism" for views that explicitly claim that reality is in some sense constructed. (rz.5).
Van Fraassen also uses the term "constructive empiricism" for his view of science, though his position has little in common with others standardly called constructivist. (I z.6)
Corroboration. Popper used this term for something that a scientific theory acquires when it survives attempts to refute it. Sometimes this looks like another name for confirmation, which Popper rejected. (4.5) The term is also occasionally used (though not by Popperians) in a way that is roughly synonymous with confirmation or support.
Covering Law Theory. A theory of scientific explanation developed by the logical empiricists (see logical empiricism), especially Carl Hempel. The theory holds that to explain something is to show how to infer it in a good logical argument that includes a statement of a law of nature in the premises. (13.2)
Deductive Logic. The well-developed branch of logic dealing with patterns of argument that have the following feature: if the premises of the argument are true, then the conclusion is guaranteed to be true. This feature is called "deductive validity."
Deductive-Nomological Theory (D-N Theory). A term sometimes used for the covering law theory of explanation, although it only refers to some of the cases covered by that theory, the ones in which the argument used to explain something is a deductive argument.
Demarcation Problem. Popper's term for the problem of distinguishing scientific theories from nonscientific theories. (4.2, 4.6)
Eliminative Inference. A pattern of inference in which a hypothesis is supported by ruling out other alternatives. (Sometimes called "eliminative induction," even though these arguments can be deductively valid in some cases.) (14.5)
Empiricism. A diverse family of philosophical views, all asserting the fundamental importance of experience in explaining knowledge, justification, and rationality. A slogan used for traditional empiricism in this book is "Experience is the only source of real knowledge about the world." Not all empiricists would like that slogan. There are also empiricist theories of language, which connect the meanings of words with experience or some kind of observational testing. (chapters z, 15, section 10.3)
Epistemology. The part of philosophy that deals with questions involving the nature of knowledge, the justification of beliefs, and rationality.
Explanandum. Whatever is being explained, in an explanation. (chapter 13).
Explanans. Whatever is doing the explaining, in an explanation. (chapter 13).
Explanatory Inference. An inference from a set of data to a hypothesis about a structure or process that would explain the data. There are many terms for this idea or ideas like it, including "abductive inference," "inference to the best explanation," "explanatory induction," and "theoretical induction."
In this book I treat this category as possibly overlapping with the category of eliminative inference. Some cases of explanatory inference might work via the elimination of alternative explanations. Others would treat these as two distinct categories. (3.z, 14.5)
Falsification ism. A view of science developed by Karl Popper. The word "falsificationism" can be used narrowly to refer to Popper's proposal for how to distinguish scientific theories from nonscientific theories (the demarcation problem). Falsificationism in this sense says that a
theory is scientific if it has the potential to be refuted by some possible observation. The term is also used more broadly for Popper's view that all testing in science has the form of trying to refute theories by observation, and there is no such thing as the confirmation of a theory by passing observational tests. (chapter 4)
Foundationalism. A term used for theories that approach epistemological problems (see epistemology) by trying to show how human knowledge is built on a "foundation" of basic and completely certain beliefs. These might be beliefs about one's own current experiences, perhaps. (ro.i, ro.z)
Holism. Holist arguments and positions can be found in many philosophical debates. Generally, a holist is someone who thinks that you cannot understand a particular thing without looking at its place in a larger whole. Two kinds of holism are important in this book. Holism about testing claims that we cannot test a single hypothesis or sentence in isolation. Instead, we can only test complex networks of claims and assumptions as wholes, because only these whole networks make definite predictions about what we should observe in a given situation. Meaning holism claims that the meaning of any word (or other expression) depends on its connections to every other expression in that language. (2.4, 2.5)
HypotheticoDeductivism. This term can be used both for a method of doing science and for a more abstract view about confirmation. The hypotheticodeductive method (H-D method) is the most common description of good scientific procedure given in science textbooks. Versions of the method vary, but the basic steps are as follows. (i) Gather some observations, (z) formulate a hypothesis that would account for the observations, (3) deduce some new observational predictions from the hypothesis, and (4) see if those predictions are true. If they are true, go back to step 3. If they are false, regard the hypothesis as falsified and go back to step z.
Some versions omit or alter step i. Versions also differ on whether the scientist should regard the theory as confirmed if the predictions made by the hypothesis are true.
"Hypotheticodeductivism" is also used for a view about the nature of confirmation, as opposed to the procedures used in testing. Here the idea is that a hypothesis is confirmed when it can be used to derive true observational predictions.
Incommensurability. An important concept in Kuhn's and Feyerabend's theories of science. The basic idea is that different theories or paradigms can be hard or impossible to compare, in a properly unbiased way. For example, incommensurability about standards is the idea that different paradigms tend to bring with them slightly different standards for what counts as good evidence or good scientific work. If two paradigms bring different standards with them, which set of standards do we use if we want to choose between the two paradigms? Incommensurability about language holds that key scientific terms (like "mass," "force," etc.) can have different meanings in different paradigms. So in a sense, people within two different paradigms can be speaking slightly different languages, even if they seem to be using the same words. (6.3)
Induction. There are many senses of this term. One old sense refers to a method for doing science described by Francis Bacon in the seventeenth century. This method is usually described as one in which lots of particular facts should be gathered first, and generalizations and other hypotheses should be based on this stock of facts. (Bacon did not think that all science should follow this simple pattern.) In most of the discussions described in this book, this is not what "induction" means. Instead, induction is a kind of argument, or pattern of inference, rather than a method or procedure.
I use "induction" for inferences in which particular cases are used to argue for a generalization that goes beyond the cases observed. So these arguments are not deductively valid (see deductive logic). The logical positivists and logical empiricists tended to use the term more broadly-for any inference that is not deductively valid but where the premises do support the conclusion to some extent. (chapters 3, 4, 14)
Instrumentalism. One kind of opposition to scientific realism. The main idea is that scientific theories should be seen as instruments used to predict observations, rather than as attempts to describe the real but hidden structures in the world that are responsible for the patterns found in observations. (r z.4, r z.6)
Likelihoods. In Bayesianism and in statistics, "likelihood" is a technical term referring to the probability that something (e) will be observed, given the truth of some hypothesis (h). So likelihoods are probabilities of the form P(elh).
The term "likely" is often not tied to this technical meaning, though. Sometimes philosophers say "likely" just to mean probable and say "likelihood" just to mean probability. (14.2, 14.3, 14.4)
Logical Empiricism. I use this term for the more moderate views about knowledge, language, and science that derived from logical positivism and developed after World War II, especially in the United States. The term is sometimes used for logical positivism too, however (especially by those who think that not that much changed between the earlier and later stages). Logical empiricism was a scientifically oriented version of empiricism that emphasized the tools of formal logic. (z.5, chapter 3)
Logical Positivism. A novel, adventurous, and scientifically oriented form of empiricism that developed between the two world wars in Vienna, Austria. Sometimes known as "logical empiricism," though I use this term for a later and more moderate development of the ideas. Leading figures were Moritz Schlick, Otto Neurath, and Rudolf Carnap. The view was based on developments in logic, philosophy of language, and philosophy of mathematics. The logical positivists famously dismissed a lot of traditional philosophy as meaningless. Early versions included the phenomenalist position (see phenomenalism) that all scientific claims could be translated into claims in a special language that referred only to observations. (chapter 2, 12-4)
Metaphysics. This term is usually now used to refer to a subfield within philosophy, which looks at a particular set of questions. These are general questions about the nature of reality itself, rather than (for example) how we know about this reality. Standard questions here include the nature of causation, the reality of the "external world," and the relation between mind and body.
The term is sometimes seen as referring to an investigation that goes beyond what can be addressed using science. Construed that way, metaphysics is regarded by many as a mistaken enterprise. (The logical positivists regarded most traditional metaphysical discussion as meaningless.) But in most current discussion, the term "metaphysics" refers to a set of questions and does not prejudge the right way to address them.
Model. A word with many senses, leading to frequent confusion. Sometimes "model" is used in science and philosophy of science just to mean a deliberately simplified theory. I generally follow another, narrower use of the term (especially in 12.7). In this sense, a model is a structure (either abstract or concrete) that is used to represent some other system. These are often, but by no means always, deliberately kept simple. The main "abstract" cases here are mathematical models used in science. In the "concrete" cases, one real physical system is used to represent another.
"Model" can also refer to an analogy that is used to accompany a theory and make it more comprehensible.
The term "model" also has a technical meaning in mathematical logic; here a model is a precise kind of interpretation of a set of sentences, one that treats the sentences as all true. This third sense has been used in philosophical attempts to formally analyze "the structure of theories," a project not discussed in this book (and about which I am skeptical).
Naturalism. An approach to philosophy that emphasizes the links (often, the "continuity") between philosophy and science. Naturalism is especially popular in epistemology and the philosophy of mind. Naturalism is sometimes taken to imply some sort of claim about the ultimately physical nature of everything that exists. So naturalists are thought to deny the existence of, for example, nonphysical souls. In this book I do not associate naturalism itself with any particular claims about what does and does not exist. For me, naturalism holds that t
he best way to address many philosophical problems is to approach them within our best current scientific picture of the world. (chapters ro, II, 15, sections 12.3, 14.5)
Normal Science. In Kuhn's theory of science, normal science is the orderly form of science guided by a paradigm. Most science, for Kuhn, is normal science. A good normal scientist applies and does not usually question the fundamental ideas supplied by the paradigm. (5.3)
Objectivity. A term often used in a vague way to refer to beliefs or beliefforming procedures that avoid prejudice, caprice, and bias. The contrast is usually with "subjective" beliefs or procedures, which bear the influence of a particular point of view.
The term is also used to refer to a way in which things can be said to exist; something exists objectively if it exists independently of thought, language, or (again) a particular point of view.
The two meanings can be combined; objectivity in the sense of lack of bias might be seen as achieved through making beliefs responsive to the real world. (1.3, 15.4)
Operationalism (Operationism). A strongly empiricist view of science and scientific language developed by a physicist, Percy Bridgman, partly in response to Einstein's work in physics. According to operationalism, all good scientific language must either refer to observations or be definable in terms that refer only to observations. So this view is similar to logical positivism but is more a suggestion for how language should be used in science than a theory of meaning applied to all language. (2.3, 8.4)