The magic — and limits — of teaching STEM through computer simulations
Almost every STEM teacher I know said a version of this last spring when we went remote: “It’s too bad we can’t do our in-person labs, but at least we have online simulations!” I shared this sentiment too, but there’s a pernicious implication: that simulations and experiments are interchangeable. They are not. I am separately drafting a longer blog post about the interplay between direct measurement, computer simulations, and reading authoritative sources in the STEM classroom. But in the meantime, since I know that STEM teachers are using simulations a lot these days, I wanted to write up some questions to guide how we teach with simulations.
What is a computer simulation? In the context of science class, “computer simulation” often refers to the likes of PHET or The Physics Aviary, both of which are go-to resources in my classroom. In a simulation, there is a computer program that mimics the real world in some key way, so that students can model how things work and make testable predictions about what would happen in certain scenarios. The computer simulation is NOT reality. When students take “measurements” in a simulation, they are not actually measuring anything physical. Rather, they are reading the output of a computer algorithm based on certain inputs. A human coded the simulation in the first place because someone else (indeed, hundreds of other people) took careful measurements and conducted mathematical calculations to derive our current understanding of how the universe works. For example, Frank McCulley, who created this lab, (probably) didn’t lift a bunch of different ramps with different blocks on it — oh yes, also on a bunch of different planets. Instead, he used known formulae to program the simulation to predict what the outcomes of the experiment would likely be.
Why use computer simulations? In the world of professional science, simulation is symbiotic with experimentation. For my college thesis, I was an experimentalist, and I collaborated with a computationalist. My collaborator developed computer models to predict the motion of colloidal particles in water, and I actually put the particles in water and imaged them. When our results didn’t match, it meant that our model for the behavior of colloidal particles was not yet useful, or alternatively that there was something going on in the experiment that we didn’t anticipate in the calculation. Computer simulations provide an additional perspective into real-world measurement.
But this isn’t how we use computer simulations in the classroom, right? Right. Well, it depends. It’s certainly possible for our students to use computer simulations to test predictions. For example, after learning about the relationship between force and displacement for a spring, I could ask students to test whether their mathematical model is consistent with this lab. But this activity does not count as verification that their model accurately represents the real world — only that it is consistent with a simulation that someone made. More commonly, though, I see teachers use simulations as if students really are taking data from the real world. Students think they’re measuring things, when they’re actually just reading outputs of a computer program. They’re not learning what “is,” but rather they’re seeing the world as scientists have modeled it.
OK, but is that a problem? It usually doesn’t matter: students collect data from online simulations, they infer a relationship between two variables, and they leave class with deeper understanding. The simulation no longer reflects reality in extreme values (that is, in the limit as quantities get really big or really small). But using a computer simulation does gloss over the inherent messiness in real-world experimentation: students falsely come to believe that their numbers will always be clean and pretty, and that instrumentation is easy to set up. They become unable to appreciate how people can spend hundreds of years debating whose description of the universe is more accurate because it’s really hard to measure the universal gravitational constant. Students think they “measured” it online in just a few seconds — when what they really did is use a simulation.
But we should still teach with simulations, right? Definitely yes. Even when this pandemic ends and we’re not teaching remotely all the time, simulations remain incredibly valuable pedagogical tools. Most of us remember doing complicated experiments in our high school science classes. By the time we had everything set up, we’d forgotten what we were doing in the first place, and then the bell rang signaling the end of class. As H. L. Terry wrote this article 100 years ago, it’s important to focus on the principles at first. Computer simulations help us do that. As I see it, there are two reasons to use computer simulations instead of direct measurement: first, when it’s infeasible to do physical experiments (because of expense, time scale, a pandemic, etc.), and second, when the hands-on experiment might distract from the key mathematical relationships that we want students to identify.
What about Pivot Interactives, or students using sensors on their phones? These are not simulations. When students do a frame-by-frame analysis of a video, or they collect data directly using sensors on their phones, they are doing direct experimentation. There is no model of how the world works from which students are “reading” the output. Rather, they are collecting actual data from the world. The phone’s camera and other sensors are used as measurement devices. They’re certainly more complicated than a ruler or even a stopwatch, but they are used to measure the world, not to predict it. Thus, not all experiments using computers are simulations. It’s only a simulation if somebody’s algorithm is being used to predict physical outcomes.
How should we modify our simulation-based lesson plans? When I use simulations in class, I do a subtle but important thing. I ask my students, “Are you measuring the real world right now?” I make sure they know that they are not taking real-world measurements. When they determine a relationship between their dependent and independent variables, that relationship is likely to bear out in the real world, but only because the person who made the simulation did so using our current understanding of how things work. Students can more flexibly engage with the mathematical relationships when they’re unencumbered by finely tuned lab supplies. And they can interact with algorithms like professionals do.
Online simulations are computer algorithms. Computer algorithms are made by people, and thus are subject to their biases. No model is complete, and no computer simulation is perfect. Encourage your students to use simulations as tools, and encourage your students to know their simulations’ limitations.