In short, models make it easy to theorize about the physical world, as seen through the eyes of experiments.

“Essentially, all models are wrong, but some are useful.” – George E.P. Box 1

This is true, if overdramatic. All of science depends on models in one way or another. Biologists have a model for how cells reproduce, physicists have models for how subatomic particles add up to atoms, and chemists have a model for how reactions make molecules. In practice, “modeling” usually refers to a specific type of activity: constructing and applying a theory which explains or predicts the outcome of a “real” experiment.

Before I explain why this is important, we have to set the scene by describing how scientists relate to each other.

Basic science research relies on the cooperative efforts of a worldwide community of scientists, engineers, teachers, students, and workers. One of the best ways to understand how basic science is practiced is to look at how the scientists sort themselves into groups. Learning about the social construction of scientific organizations has been an unexpected perk of being a graduate student.

There are two features that explain most of the organization in academia.

  1. What problem are you studying? (the question)
  2. Which method are you using? (the method)

Answering these two questions will do a good job of locating an investigator in a field. For example, I use computer simulation to answer biophysics questions about molecules that live on cell membranes. This description alone narrows the field of my peers from millions of scientists to perhaps thousands of computational biophysicists studying soft matter systems.

Unless you exclusively write science fiction, your particular method of inquiry must descend from some set of physical experiments. The difficulty with conducting scientific experiments is that the physics in most fields of inquiry are too complex to ask simple yes-or-no questions. At almost every stage of scientific inquiry, the investigator must make approximations. For example, in vitro biology experiments approximate a living organism by dividing it into parts, particle accelerators emulate conditions in outer space, and so on. Synthesis-oriented fields like chemistry and materials science tend to focus more on creating something, rather than investigating it, but their toolkit of synthesis methods relies on a number of basic models for assembling chemicals or novel materials.

If all experiments require models, it might seem redundant to claim “modeler” as an identity. However, just like microscopy or cell culture, building models can be understood as its own particular practice. Modelers describe the warp and woof of different physical experiments by testing experimental results against simplified, idealized models which seek to capture the essence of a physical problem.

In this sense, a model is a meta-experiment in which we test a theory against a natural observation and see if they match. If they do, then the model is useful. Since models are lightweight and usually easier to interrogate than experiments, they allow us to test seemingly impossibly hypotheses or make predictions about novel systems, or those that cannot be studied in a wet lab. Insofar as the model doesn’t match an experiment, it can still motivate new theory. In this sence, modeling is a complement to both theory and experiment, allowing scientists to explore the hypothesis space enencumbered by the limitations of experiments, or even physical laws themselves.


  1. Attributed to George E.P. Box. This quote is almost mandatory for self-deprecating modelers who wish to disarm angry experimenters who question their methods. More charitably, it’s a useful reminder.