Scientific Realism and Instrumentalism

Electrons are useful.

But are they real?

The same question can be asked about:

• quarks,
• fields,
• genes,
• black holes,
• spacetime curvature,
• dark matter.

Scientific theories refer to entities we often cannot observe directly.

Should we believe those entities actually exist?

Or should we treat theories only as instruments for predicting observations?

This is the debate between scientific realism and forms of anti-realism, including instrumentalism.

Scientific Realism

Scientific realism says, roughly, that our best mature scientific theories aim to describe a mind-independent world.

When such theories are highly successful, we have reason to believe that at least important parts of what they say are approximately true.

On this view:

electrons are not merely useful symbols.

They are real entities or structures in nature.

Realism Is Usually Fallibilist

Scientific realism does not require saying:

our current theories are exactly and finally true.

A realist can accept that:

• Newtonian gravity was approximate,
• classical electrodynamics has limited scope,
• current particle physics is incomplete.

The claim is weaker:

successful science often gets important structure right, even if later theories revise the details.

Instrumentalism

Instrumentalism takes a more cautious view.

A theory is an instrument for:

• organizing observations,
• predicting outcomes,
• guiding intervention.

The theory need not be interpreted as a literally true description of hidden reality.

If treating electrons as particles helps calculate detector outcomes, that may be enough.

We need not commit metaphysically to what an electron “really is.”

Prediction Without Ontology

Suppose two different mathematical models predict exactly the same observations.

If no possible experiment distinguishes them, does it matter which one is “really true”?

An instrumentalist may say no.

Science should focus on empirical adequacy.

Theories are tools.

Ontology may go beyond what evidence warrants.

Why Realists Disagree

Realists argue that this caution becomes too extreme.

Scientific theories often do more than fit known data.

They generate unexpected successful predictions.

They connect independent phenomena.

They support novel technologies.

Why would a deeply false theory work so well?

This leads to the no-miracles argument.

The No-Miracles Argument

A common realist argument says:

the success of science would be miraculous if mature theories were not at least approximately true.

Electromagnetic theory works across enormous ranges of phenomena.

Quantum theory predicts experiments to extraordinary precision.

Atomic theory explains chemistry.

The best explanation for this success, realists argue, is that theories capture genuine structure in the world.

The Pessimistic Meta-Induction

Anti-realists respond with history.

Past scientists believed in entities and theories later rejected.

Examples include:

• phlogiston,
• caloric fluid,
• luminiferous ether in its older forms.

Past successful theories were sometimes ontologically wrong.

So why assume current theories are different?

This is the pessimistic meta-induction:

history gives reason to expect current theories may also be replaced.

Realist Response

Realists answer that replacement does not erase everything.

Later theories often preserve important structure.

Newtonian mechanics remains approximately correct at low speeds and weak gravity.

Classical optics remains useful.

Old theories may contain partially correct relations even when their ontology changes.

Scientific progress can be continuity through revision.

Structural Realism

This motivates structural realism.

Perhaps science is especially good at discovering relationships and mathematical structure, even when our picture of underlying entities changes.

We may be wrong about what things intrinsically are.

We may still correctly capture how they relate.

This view tries to preserve realism while learning from theory change.

Entity Realism

Another position focuses on manipulation.

If scientists can reliably use an entity to intervene in the world, perhaps we have good reason to believe it is real.

Electrons can be produced, guided, detected, and used in technologies.

Entity realism says successful manipulation can justify belief in entities even if broader theories remain uncertain.

Constructive Empiricism

Bas van Fraassen developed an influential anti-realist position called constructive empiricism.

Science aims, on this view, at empirical adequacy.

A theory should correctly describe observable phenomena.

We need not believe its claims about unobservable entities are literally true.

Acceptance of a theory can mean using it without believing every ontological commitment.

What Counts as Observable?

This creates another question.

Observable with what?

The naked eye?

A microscope?

An electron microscope?

A particle detector?

Van Fraassen distinguishes observable in principle by unaided human capacities from merely detected through instruments, but the boundary remains philosophically controversial.

Scientific practice increasingly blurs the distinction.

Seeing Through Instruments

When we look through a telescope, do we see Jupiter?

Most people say yes.

When we use an electron microscope, do we see atoms?

The answer becomes less obvious because the image is heavily processed.

But all perception involves mediation.

The eye itself is a biological detector.

The line between direct and indirect observation is not clean.

Quarks

Quarks offer a strong test case.

They cannot be isolated individually under ordinary conditions because of confinement.

Yet quark theory explains:

• hadron structure,
• scattering experiments,
• jet production,
• particle spectra.

Should we say quarks are real?

Most working physicists do.

The inference is based on broad explanatory success rather than direct visual observation.

Black Holes

Black holes were once primarily theoretical entities.

Now multiple forms of evidence support them:

• stellar orbits,
• gravitational waves,
• accretion signatures,
• horizon-scale imaging.

At what point did black holes become “real”?

There was no magical observational threshold.

Confidence accumulated through converging evidence.

This shows realism often comes in degrees.

Dark Matter

Dark matter is more difficult.

Its gravitational effects are strongly supported.

But we do not know its microscopic identity.

A realist may believe:

something corresponding to the dark-matter component is real.

But remain agnostic about whether it is:

• a new particle,
• primordial black holes,
• something else.

Realism can attach to structure before ontology is settled.

Scientific Models Are Idealized

Realists must also handle idealization.

No gas is literally ideal.

No object is a perfect point mass.

No fluid is exactly continuous at all scales.

If successful models contain false assumptions, literal truth cannot be the only measure.

Realism therefore usually targets the parts of a theory responsible for success.

Selective Realism

Selective realism says we should believe primarily the theoretical components that are strongly involved in successful prediction and explanation.

Not every auxiliary construct deserves equal commitment.

This is an attempt to avoid both naive realism and total instrumentalism.

Underdetermination

A major challenge to realism is underdetermination.

Different theories can sometimes fit the same evidence.

If observations cannot distinguish them, which should we believe is real?

This suggests evidence may not uniquely determine ontology.

Realists respond by invoking:

• simplicity,
• coherence,
• explanatory power,
• future tests.

But the problem remains serious.

Quantum Interpretations

Quantum mechanics provides a famous case.

Several interpretations reproduce much of the same observed physics while giving very different pictures of reality.

Wavefunction collapse.

Many worlds.

Pilot waves.

Relational states.

If empirical predictions coincide, evidence alone may not choose among ontologies.

This is underdetermination at the foundations of physics.

Does Use Imply Belief?

Scientists routinely speak realistically.

“The electron entered the detector.”

“The black hole merged.”

“The gene was expressed.”

This language is practical.

Does using it commit the scientist philosophically to realism?

Not necessarily.

Scientific practice can function even when philosophical interpretation remains open.

Technology and Realism

Technology strengthens realist intuition.

Semiconductors work because quantum mechanics captures real regularities in electronic structure.

GPS requires relativistic corrections.

Lasers depend on quantum transitions.

It seems strange to say theories are merely convenient fictions when they support such precise control.

Yet instrumentalists reply:

predictive success is exactly what good instruments are supposed to provide.

The debate survives.

Approximate Truth

Realists often use the idea of approximate truth.

Newtonian mechanics is not exactly true universally.

But in ordinary regimes, it captures real relationships with extraordinary accuracy.

Science may progress by producing theories that are progressively closer to the structure of reality within wider domains.

This avoids an all-or-nothing view of truth.

What Is the Goal of Science?

The realism debate partly concerns scientific ambition.

Is the goal:

to describe what reality is like?

Or:

to construct empirically adequate systems for predicting experience?

Working science often pursues both without resolving the philosophical difference.

We want equations that work.

We also want to know what the world is.

A Balanced Position

A reasonable middle position is cautious realism.

Take mature, repeatedly successful theoretical structures seriously.

But distinguish confidence levels.

Believe more strongly in:

• well-tested relations,
• robust entities,
• converging evidence

than in speculative ontology attached to underdetermined interpretations.

Scientific belief need not be all or nothing.

The Next Problem

The realism debate becomes especially difficult when several theories explain the same evidence.

If all observable facts are compatible with multiple accounts, evidence cannot choose uniquely among them.

This is the problem of underdetermination.

What happens when evidence is not enough to determine one theory?