Rewiring Undergraduate and Middle School Science Courses: Innovation with Agent-Based Modeling

By Marilyn Sherman

In a unique approach to science courses, a SESP professor and his students use computer modeling to unlock understanding.

Doctoral student Paulo Blikstein spent three years counting the number of equations presented in an undergraduate engineering curriculum in materials science. He calculated that students were hit with an equation every one and a half minutes. He and his adviser, learning sciences associate professor Uri Wilensky, thought that there was a better way to learn the subject — and to learn it with deeper understanding.

Blikstein and Wilensky's innovative use of agent-based computer modeling in engineering classes has turned out to be a winning strategy. Computer modeling has proven equally successful in physics — not only in college courses but also with students as young as 5th grade. "Using modeling students can reason about the material rather than memorizing — it makes it possible for students to really understand," says Wilensky.

Associate professor Uri Wilensky and doctoral student Paulo Blikstein discuss the use of innovative agent-based computer modeling to improve learning in a undergraduate engineering classes
Associate professor Uri Wilensky and doctoral student Paulo Blikstein discuss the use of innovative agent-based computer modeling to improve learning in undergraduate engineering classes.
In a Chicago Public Schools cable TV show, doctoral student Pratim Sengupta explains his inventive use of computer modeling to help students as young as 5th graders learn about electricity
In a Chicago Public Schools cable TV show, doctoral student Pratim Sengupta explains his inventive use of computer modeling to help students as young as 5th graders learn about electricity.

Undergraduate engineering for tomorrow
Traditionally, the approach in undergraduate engineering classes is to provide vast numbers of equations and have students select the correct one for a given situation. But in doing so, are students really grasping how formulas represent what's going on in the real world? And with the rapid pace of scientific change, what happens in all-new situations where equations don't exist?

Using computer modeling, Blikstein redesigned a materials science curriculum "to break the divide" between undergraduate engineering students relying on equations and real-world engineers working with cutting-edge instruments and computer simulation tools. "What we're trying to do is take advanced approaches from engineering research and bring them to the classroom," he says.

The curriculum asks students to look at individual atoms, much as scientists focus in on atomic structures and their behavior, an approach that scientists are increasingly using. Students learn basic principles governing atoms, such as the fact that atoms seek to move to more stable situations. Most importantly, they apply these principles as they program their own computer models to represent scientific processes. This programming is made simple with the help of software Wilensky developed called NetLogo.

For example, take the process of a liquid transforming into a solid. To model this process, students program specific actions into each of many different computer "agents," representing the atoms. "Because students constructed the models themselves, they understand the process that generates the regularity governed by an equation," says Willensky. "They then check their model output against the textbook formula to validate their results."

"Students are actually discovering the equations by themselves instead of being told, gaining deeper understanding," explains Wilensky, who says that the breakthrough comes because constructing and running a computer model causes students to make connections that hadn't been made previously. "The usual assessments are about behavior; they don't tell you about understanding. Here, in order to get the right model, you have to demonstrate understanding." Wilensky notes.

As a further step, Wilensky and Blikstein developed a technology called "bifocal modeling" with which students can use sensors to see how their model compares with a real experiment that they run. Research shows that with this additional ability to track an actual experiment, students consistently create more precise models. Another way students can heighten their real-world understanding is by viewing their models as solid objects, using a cutting-edge three-dimensional printer.

Numerous studies show the leap in learning that occurs with agent-based modeling. In fact, Northwestern University's McCormick School of Engineering has been so impressed with the results of Wilensky and Blikstein's work that Dean Julio Ottino plans to expand the use of agent-based modeling in engineering classes.

Ottino says, "I believe that agent-based modeling in engineering will be pervasive in the future, as much as modeling things using differential equations has been in the past. Agent-based modeling is something that seems to come naturally to students who have, since birth, grown up near computers. It will expand the kinds of problems engineers can attack. I think it will be a valuable addition for our students, no matter what careers they want to pursue — either careers in the service sector or in technology."

In engineering there's a great need to train people for what's to come, especially since knowledge in the field is said to be doubling every 10 years. "What's important for engineers is adaptive expertise," says Wilensky.

Northwestern recently received a grant from the National Science Foundation to support research on how students learn engineering and how to train engineers of the future. Working with the newly established Northwestern Center for Engineering Education Research, directed by professors Ann McKenna and Rob Linsenmeier, Wilensky will lead the research for SESP.

This sequence of screen shots from a computer model was created by an undergraduate engineering student to simulate different metallic alloys. With each dot representing one atom of the metal, the model shows how adding particles of a new material (represented by the white squares) makes the resulting alloy's crystal structure smaller and therefore stronger. Traditionally, engineering students are taught this phenomenon using many formulas such as the one above, but research shows they understand more deeply by building computer models — sometimes recreating the equation by themselves. See this model in action.
(Images courtesy of Paulo Blikstein

Physics from college to kids
Using Wilensky's agent-based modeling approach, learning sciences doctoral student Pratim Sengupta has had similar success in undergraduate physics classrooms. In one study he compared physics students who used agent-based modeling with students who didn't. "The ones who used modeling did immensely better than the ones who didn't," he says.

Intrigued with the dream of teaching young children college-level physics, he spent the next year researching children's thinking about physics. He and Wilensky re-analyzed the vast literature in cognitive science about children's understanding in the area of electricity. They discovered that with agent-based modeling, they could indeed adapt and use the wealth of "folk" knowledge about electricity that young kids bring to the classroom — and this could result in powerful learning.

Their new understanding opened the door to teaching electromagnetism to middle schoolers in two Chicago public schools, one of which was Wildwood Elementary School. Using computer-based modeling, middle schoolers easily learned concepts about electricity that normally would be tackled in college.

For example, using a computer model showing electrons moving within a wire, students understood electric current and resistance. Wildwood teacher Shamiram Badal says, "NetLogo Investigations in Electromagnetism provided our students with a great opportunity to observe visually how voltage, current and resistance are related. … To explain these concepts at the age of 9 and 10 years old is extraordinary and beneficial in their future studies and everyday lives."

Wilensky is most impressed with the reasoning that emerges with modeling. "There is a huge body of literature that shows students have misconceptions about science, such as about the difference between current and voltage — and that they don't know how to reason," he says. "With computer modeling, not only are students understanding science, but they are starting to reason with it. Ideas that used to be taught at upper levels of undergraduate school now can be understood by 5th graders."

If Wilensky and his students have their way, memorizing formulas may soon be a thing of the past — and agent-based modeling the wave of the future.

See videos of computer models for engineering and electricity.
By Marilyn Sherman