Week 7- Dr.LL Guest lecture ABM

A. Complex Systems

  • used for how they affect their environment

  • Featues of Complex Systems- I

    • the development of form and patter through behaviors of individuals with no central coordination

  • Features of Complex Systems- II

    • Systems remember past events and adapt to change; systems evolve (not better, just different) over time through learning

  • Features of Complex Systems - III

    • certain behaviors can cascade across networks of indiviudals

    • Behaviors can transfer at any rate

      • after the environment has changed again the behavior rate changes

  • Reasons of importance

    • Many of today’s global challenges are a product of interactions between individuals and systems

    • Understanding these issues requires insight into the complexity of the interactions within the system

    • dynamic systems

B. Preliminiaries

  • Advantages

    • Allows appropriate modeling capabilities in a number of important disciplines

  • Applicability of ABM

    • When there are decisions and behaviors that can be well-defined

    • When it is important that agents adapt & change their behaviors

    • AGM can be multi-scalar

      • Once there is a certain threshold completed during the modeling it can be increased to run the model again

C. Agent-Based Modeling

  • Temporally and spatially AGM is important

  • Modeling Objectives of ABM - I

    • to forecast or backcast some system’s

      • behaviors of participants in the system

      • system states (micro, meso, or macro-level)

  • *Remember 3 key components of AGM: Agents, Environment, and Rules

What is an Agent?

  • Elements: autonomy, heterogeneity, active, (adaptive) learning

  • An AGM cannot run if all the agents are the same

What are rules?

  • Rules are typically based around if-then-else statements

    • these statements can be made more complicated

The Environment

  • Often spatial-2D grid of locations (easy to see what happened)

  • Sites (“patches”) can have variables to store environment state

Two types of interactions

agent-agent

  • Interactions

    • Potential neighbor relationship types of agents

      • Queen’s : Schelling’s model is the default

      • Rook’s

      • Bishop’s

      • K-Order

agent-environment

Limitations of ABM

  • There is no template or universally accepted way to design & build agent-based models

Some criticize agent-based models as being data hungry & such data does not exist

  • issues of calibration and validation

Who performs these AGM?

  • Government Agencies

  • University levels

  • Individual labs

  • All in collaboration with each other*