Thinking in Levels: A Dynamic Systems Approach to Making Sense of the World

Citation: Uri Wilensky, Mitchel Resnick (1999) Thinking in Levels: A Dynamic Systems Approach to Making Sense of the World. Journal of Science Education and Technology (RSS)
DOI (original publisher): 10.1023/A:1009421303064
Semantic Scholar (metadata): 10.1023/A:1009421303064
Sci-Hub (fulltext): 10.1023/A:1009421303064
Internet Archive Scholar (search for fulltext): Thinking in Levels: A Dynamic Systems Approach to Making Sense of the World
Download: http://ccl.northwestern.edu/papers/levels/levels.html
Tagged: Education (RSS) emergent (RSS), emergence (RSS), education (RSS), science education (RSS), levels (RSS), dynamic systems (RSS), abstraction (RSS), educational technology (RSS)

Summary

This work presents three case studies using MIT's StarLogo system, graphical educational software for simulating an emergent system. An emergent system is a system where complex large-scale behavior emerges from the interaction of simple, localized behavior of many agents, such as a traffic jam arising from interactions of individual vehicles. The goal is to demonstrate the educational benefits of approaching scientific problems as emergent systems, while also highlighting conceptual errors due to confusion between the large-scale and small-scale behavior in these systems.

The first model simulates a simple abstraction of a slime mold, designed to encourage homogeneous agents to aggregate into large clusters. SLIME highlights several counterintuitive concepts that students struggled with:

• The idea that a cluster can be viewed both as a singular entity with large-scale behavior and as a collection of agents with small-scale behavior;
• The idea that random behavior of agents is essential (rather than disruptive) to the development of large clusters;
• The idea that giving agents more information may in fact lead to smaller clusters, contradicting the intuition that they are "trying" to form large clusters.

Scientists once believed that aggregation of slime mold was organized by a differentiated leader cell rather than an emergent outcome, which the authors call the "centralized mindset": "When people see patterns in the world, they tend to assume centralized control, even if it doesn't exist."

The second case study focuses on GasLab, a simulation of an ideal gas in a box created by a teacher. The simulation reproduced and offered insight into the Maxwell-Boltzman distribution of particle speeds, and highlighted the importance of the experimenter moving continually between the large-scale and small-scale "levels," explaining one using the other. Students refined GasLab by altering temperature and volume, and developing a way to measure the resulting pressure.

The third case study involved a predatory-prey system. Whereas such systems normally require differential equations to analyze, students using these systems replicated the oscillatory populations of predators and prey over time with only simple computational rules for agents defining when birth and death occur. This also creates a more personal experience for students, who can imagine themselves in the place of individual agents.

Theoretical and Practical Relevance

The authors' main argument, that emergent systems should be taught in school to engender a more personal and accurate exploration of scientific phenomena, remains untested at a larger scale, and has not been explored in an experimental setting. In 2005, Chi found that misconceptions of emergent systems, such as those mentioned in this work, are more robust than those in direct systems ("Commonsense Conceptions of Emergent Processes: Why Some Misconceptions Are Robust").

As of 2012, MIT's StarLogo is still available and contributing to innumerable new publications in diverse areas, and the three models studied in this work are still available ("Slime", "Gas Lab Gas in a Box", and "Ecosystem - Predator, Prey, and Grass").