Archive: March 15th, 2009

Mitchel Resnick: Turtles, Termites, and Traffic Jams

[Readings] (03.15.09, 5:52 pm)

Principally, this book is about emergence and decentralization. Resnick is heavily influenced by Papert (who both was an adviser and writes the foreword). Thinking in terms of autonomous agents suggests a new paradigm of AI and pedagogy. The central foundation and observation is that things tend to organize themselves, and they are not organized by some centralized controller. Resnick explores how things organize themselves, and how to think about them.


The focus of this is decentralized systems and models. Many systems, flocks of birds, immune systems, ant colonies, market economies, and many others are decentralized. However, centralized models are pervasive, and tightly woven into our thinking. Most theories of how natural systems came to exist originated with the idea of some central control. These models are have been problematic and have often been demonstrably incorrect. Decentralized theories suggest that organized systems made of agents are composed such that the agents each have small and relatively simple rules, which when played out, tends toward organization.

Resnick suggests three points for studying these models: (p. 5)

  1. Probing people’s thinking: Investigating how people think about self-organizing behavior, and what sorts of models that people use to think about systems.
  2. Developing new conceptual tools: Coming up with heuristics and quantitative tools for thinking about decentralized systems without resorting to centralized models.
  3. Developing new computational tools: Study and test systems by building and playing with them. The substance of this is Resnick’s StarLogo.

Decentralization exists in many areas, and Resnick gives a listing of situations where decentralized systems exist and are important: organizations, technologies, scientific models, theories of self and mind, and theories of knowledge.


Resnick discusses StarLogo, a variant of Logo specifically oriented toward developing decentralized systems. StarLogo has lots of differences from Logo: there are dramatically more turtles, the turtles have senses, the space is organized in cells, these have local attributes, there are daemon processes, and means for describing rules on a general level. StarLogo includes built in commands and structures for interacting with a distributed system of agents using relatively simple instructions. The idea is to develop a pedagogically oriented approach to looking at decentralized systems. StarLogo visualizes and helps map from rules to emergent systems.

The use of construction is particularly relevant in the context of Papert’s constructionist influences. Constructionism is especially important in decentralized systems because these systems are both everywhere and tremendously misunderstood. The commonality of centralized approaches is problematic, but the reason for this is that centralized approaches are easy to understand, and we have a great deal of linguistic and conceptual tools for thinking about them. Decentralized systems are, on the contrary, unintuitive, and require simulation in order to observe and test.


Pedagogically, Resnick is interested in changing the emphasis from simulation to stimulation. He stresses thinking from the perspective of the agents within the system. He also stresses the concept of the microworld, as an experimental arena for testing ideas, rather than simulations, which are generally taken to be things based on reality. By de-emphasizing the realism, Resnick is able to open the microworlds to more freedom, openness, and experimentation. Resnick gives several examples of systems modeled by StarLogo: slime molds, ant colonies, traffic jams, and termites. The actual decentralized rules are startlingly simple. The listings of code are very short, but easily produce elegant behavior. These nonetheless suggest a significant cognitive leap from the intended system to the rules to generate that system.

Much like how Papert shows us that Logo and procedural knowledge tend to suggest an approach to mathematics that resembles calculus much more strongly than the types of math traditionally exposed to children, Resnick shows that distributed and decentralized models too lead to different models of mathematical concepts. This approach to math and geometry resembles the effects of fields and fluids, which are traditionally subjects first introduced to students in college (fluids usually late in undergraduate). That they should be so straightforward to represent using a Logo variant is nothing short of remarkable.


The centralized method of thinking is pervasive, and quickly invoked in guessing models of phenomena. It is integrated into other metaphors, language, and culture, especially in terms of leadership. This is also woven into goal and planning based models of behavior. Planning is integrally about centralized organization. In his conclusion, Resnick gives five bullets that describe characteristics of centralized models: (p. 134)

  • Positive feedback isn’t always negative. Positive feedback often plays an important role in creating and extending patterns and structures.
  • Randomness can help create order. Most people view randomness as destructive, but in some cases it actually helps make systems more orderly.
  • A flock isn’t a big bird. It is important not to confuse levels. Often people confused the behaviors of individuals and the behaviors of groups.
  • A traffic jam isn’t just a collection of cars. It is important to realize that some objects (“emergent objects”) have ever-changing composition.
  • The hills are alive. People often focus on the behaviors of individual objects, overlooking the environment that surrounds the objects.

This last point is especially noteworthy, especially in terms of cognitive models. Situation and environment are crucially important, and the emphasis on simple rules implies that cognition and decision making can effectively be pushed into the environment.

Reading Info:
Author/EditorResnick, Mitchel
TitleTurtles, Termies, and Traffic Jams: Explorations in Massively Parallel Microworlds
Tagsemergence, simulation, specials
LookupGoogle Scholar, Google Books, Amazon