Return to Video

Complexicon: Agent-Based Modeling

  • 0:00 - 0:03
    Agent-based modeling is a type of modeling in which
  • 0:03 - 0:06
    the action and interactions of autonomous agents.
  • 0:06 - 0:08
    Both with each other and the environment
  • 0:08 - 0:11
    are explicitly modeled in computer program.
  • 0:11 - 0:13
    The agents can be nearly anything,
  • 0:13 - 0:16
    animals, people, cities, nations.
  • 0:16 - 0:18
    Usually the agents follow simple rules
  • 0:18 - 0:20
    and are influenced by the other agents
  • 0:20 - 0:22
    and their immediate surroundings.
  • 0:22 - 0:25
    Let’s use an example to show what this means.
  • 0:25 - 0:27
    In this model there is some fish,
  • 0:27 - 0:29
    the agents that follow one simple rule,
  • 0:29 - 0:31
    swim straight.
  • 0:31 - 0:33
    When a fish exits on one side it,
  • 0:33 - 0:35
    re-enters on the opposite side,
  • 0:35 - 0:38
    not very excited right.
  • 0:38 - 0:39
    To make things more interesting,
  • 0:39 - 0:41
    let’s add a second rule.
  • 0:41 - 0:44
    Occasionally each fish randomly changes its direction
  • 0:44 - 0:47
    a little bit away from its current direction.
  • 0:47 - 0:50
    This is more interesting, but still not very exciting.
  • 0:50 - 0:54
    The fish aren’t interacting with each other.
  • 0:54 - 0:58
    Adding one final, third rule makes all the difference.
  • 0:58 - 1:00
    Rather than going more or less in the same direction,
  • 1:00 - 1:03
    each fish moves in more or less the average direction
  • 1:03 - 1:05
    of all the fish in its local neighborhood.
  • 1:05 - 1:08
    This mimics real life, an actual fish
  • 1:08 - 1:11
    in an actual school can’t see the entire school,
  • 1:11 - 1:14
    but it can sense the fish closed by.
  • 1:14 - 1:16
    Agent-based modeling is much more flexible
  • 1:16 - 1:18
    than other types of modeling
  • 1:18 - 1:20
    such as equation-based modeling.
  • 1:20 - 1:22
    With enough rules, almost any behavior can be modeled
  • 1:22 - 1:25
    in any phenomenon can be observed.
  • 1:25 - 1:28
    But this flexibility also has its down side.
  • 1:28 - 1:32
    Agent-based models with too many rules are hard to understand.
  • 1:32 - 1:34
    If there are twenty rules effecting one agent
  • 1:34 - 1:38
    how are we to know which rule is felt most strongly.
  • 1:38 - 1:40
    For example what rules are necessary to observe
  • 1:40 - 1:43
    a phenomenon such as the schooling of fish
  • 1:43 - 1:46
    could a simpler model yield schooling.
  • 1:46 - 1:48
    Nevertheless a well-designed agent-based model
  • 1:48 - 1:50
    can be very informative.
  • 1:50 - 1:53
    It can tell us which combinations of behavioral rules
  • 1:53 - 1:55
    for agents and environmental conditions yield
  • 1:55 - 1:58
    interesting behavior such as the schooling of fish.
  • 1:58 - 2:00
    What’s cool about this model
  • 2:00 - 2:02
    is that the fish actually forms school
  • 2:02 - 2:04
    based on three simple rules.
  • 2:04 - 2:06
    They present a different way to understand
  • 2:06 - 2:08
    how phenomena and patterns can arise
  • 2:08 - 2:09
    from very simple behavior,
  • 2:09 - 2:12
    which is a hallmark of complex systems.
  • 2:12 - 2:15
    Agent-based modeling is foundational to the science
  • 2:15 - 2:18
    that we do here at the Santa Fe Institute.
Title:
Complexicon: Agent-Based Modeling
Description:

more » « less
Video Language:
English
Duration:
02:27

English subtitles

Revisions Compare revisions