When scientists want to understand how a black hole forms or how climate will change over centuries, they can't run physical experiments. No laboratory can contain a collapsing star. No researcher can wait three hundred years for results. Yet scientists confidently make claims about these phenomena. Their tool? Computer simulations that have fundamentally changed what it means to do science.

For most of scientific history, we had two ways of knowing: theory and experiment. You either derived predictions mathematically or you tested ideas in the physical world. Simulations represent something genuinely new—a third mode of investigation that sits uneasily between the other two. Understanding their epistemic status matters because so much modern science depends on them.

Some of the most important questions in science concern systems we simply cannot manipulate. We cannot rerun the evolution of the universe with different initial conditions. We cannot deliberately warm the planet to see what happens. We cannot crash galaxies together in a laboratory. Yet these are precisely the questions that matter most for understanding our world.

Computer simulations let scientists explore the consequences of their theories in situations where physical experiments prove impossible, impractical, or unethical. A climate model implements everything we know about atmospheric physics, ocean circulation, and carbon cycles, then lets researchers ask: what happens if we double CO₂ concentrations? The simulation runs the mathematics forward through time, generating outcomes that follow necessarily from the underlying theory.

This is more than mere calculation. When scientists simulate galaxy formation, they're not just solving equations—they're creating virtual systems that exhibit emergent behaviors. The simulation becomes a kind of laboratory where phenomena can be observed, manipulated, and studied. Scientists probe their virtual galaxies much as astronomers probe real ones, discovering features that surprise even the researchers who built the models.

Takeaway

Simulations extend scientific investigation into domains where direct experiment is impossible, allowing theories to be tested against scenarios that exist only virtually.

Every simulation involves compromises. Computers cannot track every molecule in a fluid or every gravitational interaction between billions of stars. Scientists must make choices: which processes to include, which to approximate, which to ignore entirely. These decisions shape what the simulation can reveal—and what it might miss.

Consider climate modeling. A model divides the atmosphere into grid cells, but those cells can't be infinitely small. Processes smaller than the grid—like individual clouds—must be represented through simplified formulas called parameterizations. These approximations introduce uncertainty. Different parameterizations yield different results. How do we know which one captures reality?

This raises a profound epistemological question. When a simulation produces a result, is it telling us something about the world, or merely about the computational choices we made? Scientists develop sophisticated techniques—validation against known data, sensitivity analyses, comparison between independent models—to build confidence. But the gap between simulation and reality can never be closed completely. Every virtual experiment carries assumptions baked into its code.

Takeaway

Simulations are not transparent windows onto reality but constructed representations whose reliability depends on understanding which approximations matter and which don't.

Here's what makes simulations genuinely experimental: they can surprise us. Scientists sometimes discover behaviors in their simulations that no one predicted—phenomena that emerge from the complex interactions built into the model but weren't anticipated by anyone who designed it. This capacity for novelty is the hallmark of genuine experiment.

The discovery of solitons offers a famous example. In 1955, physicists Fermi, Pasta, Ulam, and Tsingou simulated a chain of particles connected by nonlinear springs, expecting to see energy distribute randomly over time. Instead, the system exhibited strange periodic behavior—energy concentrations that persisted and recurred. This computational surprise launched entirely new fields of mathematics and physics.

Such discoveries demonstrate that simulations aren't merely deductive exercises where conclusions are contained implicitly in premises. The complexity of the systems modeled means that consequences of our theories can remain hidden until simulations reveal them. When a climate model shows an unexpected feedback loop, or a protein simulation finds an unexpected folding pathway, scientists are learning something genuinely new—even if that knowledge remains indexed to the model rather than directly to nature.

Takeaway

The ability of simulations to generate unexpected results—phenomena not predicted from theory alone—gives them genuine epistemic value as tools of discovery, not just illustration.

Computer simulations occupy a fascinating middle ground in scientific methodology. They're more than mere calculation—they can surprise us, reveal emergent phenomena, and explore impossible scenarios. Yet they're less than physical experiments—their results depend on approximations, simplifications, and computational choices that may not track reality perfectly.

This isn't a weakness to be lamented but a feature to be understood. As simulation becomes ever more central to how science works, clarity about its epistemic status helps us interpret its results wisely. Simulations count as experiments in the ways that matter most: they extend our ability to investigate nature and discover truths we couldn't otherwise access.