Agent Based Modeling Presentation

Gautam Sanka
• Analyze and Elucidate the behavior of complex systems
• Complex Systems
• Collection of interconnected elements (system)
• Behavior and Characteristics cannot be anticipated from
• Any one element in system
• Sum of the elements when considered separately
• Many interrelated connections between elements
• Feedback loops, externalities, nonlinear relationships
• Paper claims that it is suited very well for Prevention Sciences
• Using it to analyze target populations looking for risk factors, various
environmental and social contexts
• Reasons to do modeling and simulation
• We think linearly- complex systems are not linear
• Constrained in that we cannot imagine and explore all possibilities in a
real system
• Cannot foresee cascading events as a result of an event
• Difficult to include Random events in mental Models
• Mental Models are too rudimentary
• Why do we Model
• INSIGHTS, not numbers.
• We want an explanation to why events can occur
• Predict Future events
• Prevention Science
• Have a number of interventions and policy options and limited resources
for implementing
• Pros/cons for each option
• Can some options work in tandem with each other
• Given X condition and Y condition
• Cigarette Example
• Gov. wants to increase tax on cigarette- Implications?
• Black Market trade
• Usage of other tobacco products
• Unhealthy dependence on tax revenue (unstable revenue)
more
descriptive
• “Accounting” and Data Models
• Statistical Modeling, Inductive
Inferencing (Data Driven Models)
• Social Network Analysis (SNA)
• Systems Dynamics (SD)
more process
oriented
• Agent-based Modeling/Complexity
(ABMS)
• Relationships between individuals, groups, agencies,
geographical locations
• Nodes are the groups and links are relationships
• Centrality and see hidden networks
Known
Unknown
• Most applied model in prevention research
• Example Project TND (Towards No Drug Abuse)
• Reduced youth substance use in short and long term
• Decades of Research-peers provide critical context
• To test,
• TND Networked
• Students wrote Five Best friends, best person for group leader and this
helped create a network on the computer
• Score developed to indicate substance use among each participant’s
friend network
• TND Network were less likely to sue substances compared to
controls
• Youths with higher levels of substance use among friends were
more likely to increase substance usage
• Ground Breaking research as proved that social networks were
active elements in prevention efforts
• Aggregates individual entities and continuous quantities into
specific groups
• Simulation is prepared
• Exploration of questions about why systems behave the way
they do and helps identify leverage points
• Tools
• Casual loop Diagrams- casual relationships
• Stock and Flow Models- simulate accumulations within a system over time
Media
State
Terrorists
??
Public Opinion
· Necessity
· Legitimacy
Cultural
Resources
Opinion
Leaders
Suicide Terrorists
Ideology
Culture of
Martyrdom
Level of
Grievance
Occupation
Policy
Population
(non-terrorists)
Level (Stock)
Rate
Auxiliary Variables
• SD simulations consist of equations that can be solved forward
in time:
Statet+1 = Statet + Ratet
where Ratet = f(Statet-1, … ,State0)
• Drawbacks
• Macro-model of a system
• Qualitative approach
• Many variables that cannot be quantified
• An agent is
• An individual with a set of attributes or characteristics
• Placed in an artificial environment
• A set of rules governing agent behaviors is made
• Responds to the environment
• Interacts with other agents
• Rich quantitative methodology that explores how certain
components give rise to multi-layered phenomena
• Can be used for many applications ranging from anthropology
to health
• Heroin Effects in Denver (consumer, producer, distributor)
• Simulated roles, motives, behaviors and interactions of market participants
• Consumers and brokers are most complex
• Heroin addiction changed based on heroin usage, past experiences,
transaction partners
• Police and homeless people-less complex
• Typical market conditions and reactions
COPYRIGHT 2005 SCIENTIFIC AMERICAN, INC.
• Environmental conditions are
put in the systemprecipitation levels, ground
water locations, climatic
shifts
• Simulation is pretty
accurate although the
settlements are not as
precise
Objects in the TAP Model
Objects learn and adapt based on their
history, current state, and the states of other
objects.
Simulation is built upon many different
instances of these object types, each with
different attributes.
The object architecture allows for flexibility:
the PersonRole class, and its inherited
subclasses, allow a construct where any one
Person object can play multiple roles.
Interfaces allow for specification of
required actions that can be implemented
differently, depending upon the type of
object implementing the interface
• Model must be verified and validated
• Verification
• Verifying that the model does what it is intended to do from an
operational perspective
• Validation
• Validating that the model meets its intended requirements in terms of the
methods employed and the results obtained
• Questions?