What Are Scientists Really Trying to Do?

Imagine you’re in science class, and your teacher asks a question: “What is science for?” You might say something like, “It’s to figure out how the world works.” And that seems right. Scientists build theories about atoms, genes, black holes, and quarks. They make predictions, run experiments, and slowly piece together a picture of reality. But here’s a strange question: when scientists say a theory is “true,” what exactly are they claiming?

Think about it this way. You can see a tree. You can touch it. You know it exists. But can you see an electron? Can you touch a quark? No one has ever directly observed a single electron. Scientists believe in electrons because their theories say electrons exist, and those theories work really well at predicting what we can observe. But is believing in something you can’t see the same kind of belief as believing in something you can?

This is where a fascinating debate begins. One group of philosophers, called scientific realists, says: yes, science aims to give us a literally true story of everything—the observable and the unobservable alike. When you accept a scientific theory, you should believe it’s true, all the way down.

But another philosopher named Bas van Fraassen offered a different view in 1980. He called it constructive empiricism. And his idea is both simpler and weirder than it sounds.

The Aim of Science: A Surprising Claim

Van Fraassen says that science aims to give us theories that are true about the observable parts of the world—but it doesn’t aim at truth about unobservable parts. That’s the “empiricism” part: we should only commit ourselves to what we can experience. The “constructive” part means that scientists are actively constructing theories that organize and predict observable phenomena.

If you accept a theory as a constructive empiricist, you believe it’s “empirically adequate.” That’s a technical term, but here’s what it means: the theory correctly describes everything we can observe. Past observations, future observations, observations that could be made if we were in the right place at the right time. But as for the unobservable stuff the theory talks about—the electrons, the quarks, the dark matter—you’re allowed to be agnostic. You don’t have to believe they exist, and you don’t have to believe they don’t. You just don’t know.

This might sound like a weird, weak position. But van Fraassen thinks it captures what science actually does better than the realist view does.

An Important Distinction: Aims vs. Beliefs

Here’s something easy to misunderstand. Constructive empiricism is a view about what science aims to do, not a rule about what individual scientists must believe. Van Fraassen makes a useful distinction:

A scientific gnostic believes the science they accept is true (both observable and unobservable parts).
A scientific agnostic believes the science they accept is empirically adequate, but doesn’t commit to the unobservable parts.

You could be a scientific gnostic and still be a constructive empiricist—you’d just be choosing to believe more than science itself requires. The constructive empiricist says: science as an enterprise aims at empirical adequacy. What individual scientists believe beyond that is their own business.

Think of it like a game. The aim of chess is to checkmate your opponent. That’s what counts as success in chess. But individual players might have personal aims—to have fun, to get better, to impress their friends. The aim of the game isn’t the same as the aims of the players. Same with science.

What Does “Empirically Adequate” Mean Exactly?

Van Fraassen doesn’t think of scientific theories as collections of sentences in a particular language. He thinks of them as models—structures that describe how things could be. A theory is a family of these models.

A theory is empirically adequate if the observable phenomena we actually encounter can “fit into” one of the theory’s models. Van Fraassen says the observable phenomena must be isomorphic to the “empirical substructures” of some model. That’s a fancy way of saying: the patterns we see in the observable world can be matched up with a part of the theory’s structure.

Here’s a concrete example. Suppose you have a theory about invisible particles bumping into each other. You can’t see the particles, but you can predict what will happen when you mix certain chemicals. If your predictions keep coming true, the theory is empirically adequate—the observable results fit the model. You don’t need to believe the invisible particles are really there for the theory to work.

This is important because it helps solve a problem. Someone might say: “If sentences about observables are logically connected to sentences about unobservables, then believing the observables commits you to the unobservables.” But van Fraassen’s model-based approach avoids this. Believing that observable phenomena can be embedded in a theoretical model doesn’t require you to believe the unobservable parts of that model are real. You’re just saying the pattern fits.

What Counts as Observable?

This is where things get tricky. Van Fraassen says observation must be unaided—no microscopes, no telescopes, no instruments that extend your senses. You observe things directly with your senses, the way you see a tree or hear a bird.

“Observable” doesn’t mean “currently being observed.” It means something could be observed under the right circumstances. The moons of Jupiter are observable—if you were close enough, you’d see them. But a virus seen through an electron microscope? According to van Fraassen, you’re not observing the virus itself. You’re observing an image produced by the microscope.

This distinction matters because it draws a line between what science can claim to know and what it can only hypothesize. Many philosophers have objected to this line. They say it’s arbitrary—why should a glass lens count as “aiding” perception when your own eyeball is a kind of natural lens? Van Fraassen’s response is that our human senses are the foundation of all empirical knowledge. Extending them with instruments gives us useful information, but it doesn’t give us the same kind of direct contact with reality.

What Does It Mean to “Accept” a Theory?

Acceptance is more than just belief. When you accept a theory, you believe it’s empirically adequate. But you also make a commitment: you promise to use the theory as a framework for further research, to answer questions within its terms, and to see if new phenomena can be accommodated without giving up the theory.

Acceptance comes in degrees. A scientist might strongly accept a theory, using it as the basis for their work, without fully believing it’s empirically adequate in every detail. They might think most of the observable phenomena fit, and they might be committed to the research program, but they’re not all-in.

This helps explain something you’ve probably noticed: scientists often talk as if their theories are literally true. They say “electrons have negative charge” the same way they say “water boils at 100°C.” The constructive empiricist says scientists are speaking from within the theory, “conceptually immersed” in its language. They’re not necessarily expressing full belief—they’re doing their job as explainers and researchers.

Why Would Anyone Be a Constructive Empiricist?

1. Less Epistemic Risk

Believing a theory is empirically adequate is a weaker claim than believing it’s true. Both claims go beyond the evidence—we can’t check every possible observation. But as van Fraassen puts it: “It is not an epistemological principle that one might as well hang for a sheep as for a lamb.” If you can be right with less commitment, why commit more?

2. It Makes Sense of Scientific Practice

If you watch what scientists actually do, they spend most of their time designing experiments, collecting data, and looking for patterns in observable phenomena. They’re not primarily concerned with whether their theories are true in some ultimate sense—they’re concerned with whether they work. Van Fraassen argues that the real importance of theories is that they guide experimental design. Scientists need theories to figure out what to look for and how to look for it.

3. Pragmatic Virtues Aren’t Truth Indicators

Scientists value theories that are simple, elegant, mathematically beautiful, and explanatory. But are these features signs of truth? Van Fraassen thinks not. He points out that there’s no reason to think the universe is simple rather than complicated. Some of the most empirically successful theories (like the Standard Model of particle physics) are famously messy. Meanwhile, beautiful unified theories (like Kaluza-Klein theory) have failed.

One study found that people prefer simple explanations even when the more complex explanation is ten times more likely to be correct. Our preference for simplicity seems to be a psychological fact about us, not a sign that simple theories are truer.

4. Avoiding “Inflationary Metaphysics”

If you believe a theory is true, you’re committed to believing in all sorts of things: laws of nature, natural kinds, objective necessity, maybe even possible worlds. Van Fraassen calls this “inflationary metaphysics.” The constructive empiricist can avoid all that. She can use language about what’s possible or necessary as talk about what fits in the theory’s models, not as claims about the deep structure of reality.

Arguments Against Constructive Empiricism

The Miracle Argument

Hilary Putnam famously argued that scientific realism is “the only philosophy that doesn’t make the success of science a miracle.” If our theories aren’t true, why do they work so well? Isn’t it an amazing coincidence that false theories keep making correct predictions?

Van Fraassen’s response is clever. He points to evolution: only the successful theories survive. The scientific landscape is “a jungle red in tooth and claw.” Theories that don’t latch onto actual regularities in nature get discarded. So it’s no miracle that the theories we have are successful—it’s just natural selection for ideas.

The Observable/Unobservable Line

Many philosophers argue that the distinction between observable and unobservable is vague, arbitrary, or impossible to draw cleanly. Paul Churchland pointed out that it’s just a contingent fact about humans that we can change our location but not our size. If we could shrink ourselves, we’d “observe” microbes directly. So why treat size as the important factor?

Van Fraassen’s reply: the distinction isn’t based on some metaphysical difference in the world. It’s a distinction about what we can experience. And since experience is our only source of information about the world, it makes sense to treat it as special.

Instrument-Aided Observation

Ian Hacking made a powerful argument. Consider a machine that makes tiny metal grids. You can see the large grids with your eyes. You check the smaller ones with a microscope. You can verify the process works at every scale. Eventually the machine makes grids so small you can only see them through a microscope. But you know the process is reliable. Hacking says it would be unreasonable to doubt that you’re really seeing the tiny grid through the microscope.

Van Fraassen responds: you’re assuming the grid was successfully made. That’s exactly what’s in dispute. And the fact that different microscopes agree doesn’t prove anything—the constructive empiricist can simply be agnostic about whether there’s a real grid there, just as we’re agnostic about the reality of rainbows as physical objects.

Is Constructive Empiricism Too Modest?

One powerful objection is that constructive empiricism isn’t empiricist enough. If you really want to be faithful to experience, why believe a theory is adequate for all observable phenomena—past, present, and future? That goes way beyond what anyone has actually observed. Shouldn’t a true empiricist only believe what’s been observed so far?

Van Fraassen’s reply is that science wouldn’t make sense if its aim were only truth about what’s already been observed. Scientists do experiments to discover new phenomena. They push beyond the limits of current observation. If the aim were just “truth about what we’ve seen,” there would be no point in doing new experiments. This shows that the aim of science has to include the observable-but-not-yet-observed.

The Bottom Line

Constructive empiricism is an attempt to be a scientific anti-realist without rejecting science itself. It says: science gives us wonderful, powerful theories that let us predict and control the observable world. But it doesn’t give us knowledge about unobservable reality. We can use those theories, admire them, and build on them, without committing ourselves to believing every entity they postulate.

This is still an active debate. Philosophers argue about where to draw the observable/unobservable line, whether the concept of empirical adequacy is coherent, and whether constructive empiricism really makes better sense of scientific practice than realism does. Nobody has settled it.

But the core question is one you can think about yourself: When you learn about electrons and quarks and black holes in science class, do you believe they exist? Or do you just believe the theory works? And is there even a difference?

Key Terms

Term	What it does in this debate
Constructive empiricism	The view that science aims at truth about observable things, not about unobservable things
Scientific realism	The view that science aims at truth about everything, including unobservable things
Empirical adequacy	A property a theory has when everything it says about observable things is true
Observation (unaided)	Perceiving something directly with your senses, without instruments
Acceptance	The combination of believing a theory is empirically adequate and committing to use it for research
Pragmatic virtues	Features like simplicity and explanatory power that make a theory appealing but don’t show it’s true
Inflationary metaphysics	The extra commitments (like laws of nature, possible worlds, objective necessity) that come with believing a theory is fully true

Key People

Bas van Fraassen (born 1941): The philosopher who created constructive empiricism in his 1980 book The Scientific Image. He wanted to show how someone could be an empiricist—someone who only trusts experience—and still fully support science.
Hilary Putnam (1926–2016): A philosopher who argued for scientific realism using the “miracle argument”—that if theories aren’t true, their success is a miracle. Van Fraassen disagreed.
Paul Churchland (born 1942): A philosopher who criticized constructive empiricism by arguing that the observable/unobservable distinction is arbitrary.
Ian Hacking (born 1936): A philosopher who argued that we can see things through microscopes and should believe in their reality, contrary to constructive empiricism.

Things to Think About

Think of a scientific theory you’ve learned about—atomic theory, for example. Can you separate what it says about observable things (chemical reactions, states of matter) from what it says about unobservable things (electrons, protons, quarks)? Does believing the first part really not commit you to the second?
If you were a constructive empiricist, would you behave differently as a scientist than a scientific realist would? Would you design different experiments? Interpret results differently? Or does the view only matter for what you believe, not for what you do?
Van Fraassen says we should only trust unaided observation. But what counts as “unaided”? Are glasses “aids”? Eyeglasses correct your vision—are you observing directly with or without them? Contact lenses? Lasik surgery? Where do you draw the line, and why?
The constructive empiricist says we can be agnostic about unobservable entities. But many unobservable entities (like electrons) are central to theories that make incredibly precise predictions. At what point does agnosticism start to seem like stubbornness? When, if ever, does it become unreasonable to doubt?

Where This Shows Up

Science class debates: When teachers say “we know atoms exist,” are they stating a fact or expressing a philosophical position? Different science textbooks present realism differently.
Public debates about science: Arguments about whether to “believe in” climate change or evolution often turn on whether scientific consensus about unobservable processes should count as knowledge.
Everyday reasoning: When you use a computer or a phone, you’re relying on theories about electrons and semiconductors. Do you have to believe those theories are true to use the technology?
Medicine: Discussions about whether we should believe in germs (unobservable without microscopes) based on their observable effects (disease, infection) mirror the realist/anti-realist debate.