Seminar: Steve Railsback

March 23, 2006, 4pm
335 West Hall
University of Michigan

Steve Railsback
Lang, Railsback & Associates and Humboldt State University
Arcata, California

"Individual Adaptive Behavior in Models of Complex Systems"

In real natural and economic systems, system properties emerge from adaptive individual behavioródecisions made by individuals in response to changes in themselves and their environment, presumably to improve the individual's future success. When we try to understand a complex system by modeling it, we must decide which adaptive behaviors need to be included in the model and how to model those behaviors. Our experience with individual-based models for management ecology (see: www.humboldt.edu/~ecomodel) indicates that models with few or no adaptive behaviors tend to be uninteresting and poorly able to reproduce basic system behaviors, but models with too many adaptive behaviors can be difficult to build and validate. Several approaches to modeling adaptive behavior are widely used. Economists seem particularly fond of using adaptive computation to artificially evolve decision-making traits. Ecologists have often used heuristics or logical rules. We have had success with a third approach: explicit fitness-seeking. An explicit measure of expected future fitness is defined; in ecology this measure could be as simple as growth rate or as complex as a prediction of future reproductive success. (Translated to economics, an explicit measure of individual utility could range from income rate to expected probability of attaining a future goal such as comfortable retirement.) Decisions are then made by selecting the alternative that maximizes the fitness measure. If the fitness measure is well-designed and the model thoroughly analyzed, this approach can give model individuals good (yet realistically bounded) decision-making ability over a wide range of alternatives and conditions, and also provide mechanistic understanding of how individuals make decisions and how decision-making traits affect the system.

Our work on stream trout illustrates these ideas. We developed and tested models for two key adaptive behaviors: habitat selection (choosing which habitat patch to occupy) and activity selection (deciding whether to feed or hide). First, we determined that conventional theory of behavioral ecology was too simplistic to be useful in a model with realistic variation over time and space and among individuals. We then developed appropriate fitness measures and conducted extensive analyses to show that they provide trout with the ability to make good tradeoffs between growth and mortality risk over very diverse conditions. Currently we are using the model to investigate the relative importance of these two adaptive behaviors to population properties such as abundance, variability, resilience, and resistance to disturbance.

[Home] [Syllabus] [Classnotes] [Resources] [Projects]

[Created by Greg Madey: gmadey@nd.edu]   [Notre Dame Home]   [COE Home]   [CSE Home]