Chapter 14: Instrumentalism & Constructive Empiricism

It is crucial to recognize that anti-realism is not a monolithic position. Different anti-realists deny different realist commitments. Some deny the semantic thesis (instrumentalists claim theories are not genuine descriptions). Some deny the epistemic thesis (constructive empiricists accept theories as literal but refuse to believe their claims about unobservables). Some question the metaphysical thesis (social constructivists question whether there is a mind-independent structure that science discovers). And some deny that the debate itself is well-posed (Fine’s NOA). Understanding which commitment is being challenged is essential to evaluating each anti-realist position.

Instrumentalism holds that scientific theories are not descriptions of reality but instruments — tools for organizing observations and generating predictions. On this view, it is a category mistake to ask whether a theory is “true.” The appropriate evaluative question is whether the theory is useful, accurate, fruitful, or elegant.

Pierre Duhem articulated a sophisticated version of this position in The Aim and Structure of Physical Theory (1906). Duhem distinguished between explanation and representation. Theories, he argued, do not explain phenomena by revealing their underlying causes; they represent or classify experimental laws in an economical manner. The aim of physical theory is to provide a “natural classification” of experimental laws — nothing more.

“A physical theory is not an explanation. It is a system of mathematical propositions, deduced from a small number of principles, which aim to represent as simply, as completely, and as exactly as possible a set of experimental laws.”

Ernst Mach went further, arguing that science should aim at the “economy of thought.” Theoretical concepts like “atom” and “force” are intellectual conveniences that help us organize sensory experience efficiently. Mach famously resisted the atomic hypothesis, insisting that atoms were metaphysical additions to what could be described in purely phenomenal terms. He was reportedly unconvinced even by Perrin’s experiments, though he softened his opposition near the end of his life.

The logical positivists inherited this instrumentalist spirit. Rudolf Carnap proposed that theoretical terms get their meaning through “correspondence rules” that connect them to observation terms. Without such rules, theoretical statements are meaningless. This view made the theoretical/observational distinction philosophically fundamental — a distinction that later anti-realists would redraw in different ways.

In 1980, Bas van Fraassen published The Scientific Image, a book that transformed the realism debate. Van Fraassen rejected both naive instrumentalism and scientific realism, offering instead a position he called constructive empiricism. The central thesis is concise and powerful:

“Science aims to give us theories which are empirically adequate; and acceptance of a theory involves as belief only that it is empirically adequate.”

Empirical Adequacy vs Literal Truth

A theory is empirically adequate if what it says about the observable things and events in the world is true. Crucially, van Fraassen accepts that theories should be interpreted literally — they make genuine claims about unobservable entities. He is not an instrumentalist who denies that “electrons exist” is a meaningful claim. He simply holds that we need not believe such claims. Accepting a theory means believing it is empirically adequate, not that it is true.

The distinction between acceptance and belief is crucial. When a scientist accepts a theory, she commits to using it as a framework for research — designing experiments within its terms, pursuing its research program, explaining phenomena by its lights. But this pragmatic commitment does not require believing the theory’s claims about unobservable reality. One can be a productive scientist while remaining agnostic about electrons, quarks, and dark matter.

The Observable/Unobservable Distinction

Van Fraassen’s position depends on a principled distinction between the observable and the unobservable. He defines observability in terms of the capacities of human beings as biological organisms. Something is observable if there are circumstances in which we could observe it with our unaided senses. The moons of Jupiter are observable (we could travel there and see them); electrons are not.

This draws a line using a criterion internal to science itself — what science tells us about human perceptual capacities determines what counts as observable. Critics have attacked this from multiple angles:

Van Fraassen responds that the observable/unobservable distinction need not be sharp to be philosophically significant. Even a vague boundary is a real boundary. And the anthropocentrism objection misses the point: the question is not what is “real” but what we have sufficient epistemic grounds to believe. Our direct perceptual access gives us stronger grounds for belief about observables than about unobservables.

Van Fraassen’s account of explanation is central to his anti-realism. The realist appeals to inference to the best explanation (IBE) as a ground for believing in unobservable entities: the best explanation of certain phenomena involves positing electrons, so we should believe electrons exist. Van Fraassen undercuts this by arguing that explanation is pragmatic rather than objective.

On van Fraassen’s view, an explanation is an answer to a why-question, and what counts as a good answer depends on the context — on the contrast class (what alternatives are being compared), on the relevance relation (what kinds of factors are being sought), and on the background knowledge of the questioner. There is no context-free, objective relation of “being the best explanation.”

This pragmatic theory of explanation neutralizes the realist’s appeal to IBE. If explanatory goodness is contextual and pragmatic, it cannot serve as a guide to truth about unobservable reality. A theory can be the best explanation in a given context — the most salient, the most informative, the most unified — without being true. The virtues that make explanations good (simplicity, elegance, unifying power) are pragmatic virtues, not indicators of truth.

Arthur Fine offers a provocative deflationary alternative to both realism and anti-realism. In his influential papers “The Natural Ontological Attitude” (1984) and “And Not Anti-Realism Either” (1984), Fine argues that both positions share a common mistake: they add a philosophical gloss to what science tells us that science itself does not warrant.

Fine’s NOA begins with what he calls the “core position” — the minimal stance shared by both realists and anti-realists. Both accept the results of scientific investigation as “true” in the ordinary, homely sense. When a scientist says “there are electrons,” both the realist and the anti-realist accept this as a legitimate scientific result. Where they diverge is in the philosophical interpretation they impose:

Fine’s radical suggestion is that we should stop adding. NOA accepts the results of science at face value, using “true” in its ordinary, deflationary sense, without inflating it into a grand metaphysical correspondence or deflating it into mere empirical adequacy. Scientists say electrons exist; we should take this at face value and resist the philosopher’s temptation to ask “but do they really exist?”

“NOA suggests that the legitimate features of realism (and anti-realism) are already contained in the scientific practices that both camps endorse. All the extras are just philosophical baggage.”

Critics respond that NOA is either unstable — collapsing into realism or anti-realism depending on how “truth” is understood — or too thin to be a genuine philosophical position. Musgrave has argued that if NOA takes “true” in its ordinary correspondence sense, it just is realism; if it deflates truth, it abandons the metaphysical commitments that scientists themselves seem to endorse.

The debate over the observable/unobservable distinction has continued to evolve. Robert Shapere argued that “direct observation” through instruments (such as observing the sun’s interior through neutrino detection) should count as genuine observation, not mere theoretical inference. Dudley Shapere proposed that what counts as “observation” changes as science develops — it is a historically evolving concept, not a fixed philosophical category.

More recently, philosophers have turned to cognitive science and the philosophy of perception for help. If perception itself is theory-laden (as the cognitive penetrability thesis suggests), then the distinction between direct observation and theoretical inference may be less sharp than van Fraassen assumes. On the other hand, if there is a meaningful distinction between perceptual experience and cognitive judgment, this might support a principled (if vague) boundary between the observable and the unobservable.

A more radical anti-realist tradition emerges from the sociology of scientific knowledge (SSK). Social constructivists argue that scientific “facts” and “discoveries” are not simply read off from nature but are constructed through social processes — negotiation, persuasion, institutional power, and cultural context.

The Edinburgh School (Strong Programme)

David Bloor and Barry Barnes developed the “Strong Programme” in the sociology of knowledge, articulating four principles: causality (social causes explain beliefs), impartiality (explain true and false beliefs alike), symmetry (use the same types of cause for both), and reflexivity (the programme applies to itself). The crucial principle is symmetry: the sociologist should explain why scientists accept a theory using the same types of social explanation, regardless of whether the theory is “true” or “false.”

Latour and Woolgar: Laboratory Life

Bruno Latour and Steve Woolgar’s Laboratory Life: The Construction of Scientific Facts (1979) provided an ethnographic study of the Salk Institute, observing scientists as an anthropologist would observe a remote tribe. They argued that scientific facts are “constructed” through a process of inscription and literary persuasion. A scientific fact is not discovered in nature; it is the end product of a process in which statements gradually lose their qualification markers (from “it is possible that X” to “X has been shown to be the case” to simply “X”).

“A fact is merely a statement with no modality… and no trace of authorship.”

Latour later developed Actor-Network Theory (ANT), arguing that human and non-human “actants” (instruments, microbes, equations) form networks that produce what we call scientific knowledge. On this view, the distinction between “nature” and “society” is itself a product of the network rather than a pre-given framework.

Responses and the Science Wars

Social constructivism provoked fierce backlash, culminating in the “Science Wars” of the 1990s. Alan Sokal’s hoax paper in Social Text (1996) brought the debate to public attention. Realists argued that constructivists confuse the context of discovery (how scientists come to believe something) with the context of justification (whether the belief is warranted). Social factors may influence which problems scientists investigate and how they frame results, but the truth of scientific claims is determined by evidence, not by social negotiation. The constructivist responds that even “evidence” is theory-laden and socially mediated.

Van Fraassen’s constructive empiricism is part of a broader empiricist program that extends to his views on modality, probability, and the limits of knowledge. In Laws and Symmetry (1989), van Fraassen argues against the reality of laws of nature, proposing instead that what we call “laws” are features of our models rather than features of the world. In The Empirical Stance (2002), he characterizes empiricism not as a thesis (a set of beliefs) but as a stance — an attitude toward inquiry characterized by suspicion of metaphysical explanations and commitment to grounding knowledge in experience.

This stance-based approach answers a long-standing objection to empiricism: if empiricism is itself a thesis, it seems to be a metaphysical thesis — and hence self-undermining. By treating empiricism as a stance rather than a thesis, van Fraassen avoids this self-refutation. The empiricist stance involves commitments to certain policies and practices (seeking empirical evidence, resisting metaphysical inflation) rather than to specific metaphysical claims.

Van Fraassen’s voluntarism about rational belief further supports his anti-realism. He argues that rationality does not compel belief in the truth of our best theories; it merely requires that our beliefs be internally coherent. Two rational agents can look at the same evidence and legitimately disagree about whether to believe a theory is true or merely empirically adequate. The realist’s claim that rationality demands belief in truth is, van Fraassen argues, an unwarranted inflation of the requirements of rationality.

Contemporary anti-realism has developed well beyond the classical positions. Kyle Stanford’s problem of unconceived alternatives represents a particularly influential recent argument. Stanford argues that the history of science reveals a systematic pattern: scientists have repeatedly failed to conceive of the theories that would eventually replace their current best theories. Nineteenth-century chemists could not conceive of quantum mechanics; Newtonians could not conceive of general relativity. This suggests that we too are probably failing to conceive of the theories that will eventually replace our own.

Stanford’s “new induction” differs from Laudan’s pessimistic meta-induction in an important respect. Laudan argues from the falsity of past theories to the probable falsity of current ones. Stanford argues from our historical inability to conceive of genuine alternatives to the probable existence of currently unconceived alternatives. Even if our current theories are approximately true, we may be unable to identify which parts are true, because the true alternative theory may organize the evidence in a way we cannot currently imagine.

Structural empiricism, developed by Otavío Bueno, combines van Fraassen’s empiricism with structuralist insights. Bueno argues that we should be empiricists about both observable and structural claims — accepting structural descriptions as empirically adequate without committing to their truth. This provides a more nuanced anti-realist position that can accommodate the structural continuity through theory change without conceding the realist’s conclusions.

A common objection to anti-realism is that it fails to account for actual scientific practice. Scientists routinely talk as if their theories describe real entities and processes. They design experiments to detect particles, measure properties, and test hypotheses about unobservable phenomena. This practice seems to presuppose realism.

Van Fraassen responds that the constructive empiricist can fully participate in scientific practice without endorsing realism. A scientist can design experiments, make predictions, and explain phenomena while “accepting” theories as empirically adequate rather than “believing” them true. The internal discourse of science (“electrons have charge -e”) need not be taken as expressing metaphysical commitments; it can be understood as talk within an accepted framework.

However, critics argue this picture is psychologically unrealistic. Scientists who devote their careers to studying neutrinos or dark matter typically believe these entities exist. The realist argues that this belief is epistemically warranted and that anti-realism, whatever its philosophical merits, is a position adopted by philosophers about science rather than by scientists themselves. Van Fraassen might reply that the philosophical status of a position does not depend on whether scientists endorse it — scientists are not necessarily the best philosophers of science.

Van Fraassen, The Scientific Image (1980)

This book transformed the realism debate by showing that a sophisticated anti-realism could be articulated without reverting to the discredited positivist program. Van Fraassen accepts that theories should be taken literally (contra instrumentalism) but insists that acceptance of a theory requires only belief in its empirical adequacy, not its truth. The book also develops a pragmatic theory of explanation that undercuts the realist’s use of inference to the best explanation.

Fine, “The Natural Ontological Attitude” (1984)

Fine’s twin papers argue that both realism and anti-realism are philosophical add-ons to the “core position” that both share. The realist inflates truth into a metaphysical correspondence; the anti-realist deflates it into mere empirical adequacy. NOA says: stop adding. Take the results of science at face value, using “true” in its ordinary sense, and resist the temptation to philosophize further.

Latour & Woolgar, Laboratory Life (1979)

This ethnographic study of laboratory practices argued that scientific facts are “constructed” through social processes of inscription, negotiation, and persuasion. The book helped establish the sociology of scientific knowledge as a field and provoked intense debate about the social dimensions of science.

Stanford, Exceeding Our Grasp (2006)

Stanford develops the “problem of unconceived alternatives” as a new challenge to scientific realism, distinct from both the underdetermination argument and the pessimistic meta-induction. He argues that our systematic failure to conceive of the theories that will eventually replace our current ones undermines the realist’s claim that our best theories are approximately true.

Anti-realist positions can be arranged on a spectrum from modest to radical:

Most contemporary anti-realists occupy positions near the modest end of the spectrum. Constructive empiricism remains the most widely discussed and philosophically sophisticated form of anti-realism. The more radical positions — particularly strong social constructivism and relativism — are held by a minority and face serious philosophical objections, including charges of self-refutation and inability to account for the manifest success of science.