ITS 832
Chapter 13
Management of Complex Systems: Toward Agent-Based Gaming for Policy
Information Technology in a Global Economy
Professor Michael Solomon
1
Introduction
Simulating/Managing Social Complex Phenomena
Leadership and Management in Complex Systems
Serious Gaming
Agent-Based Games for Testing Leadership and Management
Single and Multiplayer Settings
Summary and conclusions
Simulating and Managing Social Complex Phenomena
Study of how people interact
Scale prohibits experimentation with real populations
Agent-Base modeling (ABM)
Networked agents
Each agent is an individual
Interaction may modify agent behavior
Managing complex phenomena introduces complexity
Techniques to manage turbulent situations vary
Technique success depends on responding to agent behavior
Which may change based on interactions
Leadership and Management in Complex Systems
Traditional leadership research
Generally focuses on single period in time
Doesn’t address dynamic relationships
Timing of leadership principle application matters
Primary leadership functions
Instructional and regulatory
Developmental
Simulations offer promise to help model leadership in complex systems
Serious Gaming
Applying gaming techniques to real life situations
Flight simulators
Effective for evaluating complex environments
Player must interact with multiple actors and situations
Currently used for side range of training applications
Leadership use
Deterministic – limited scope
ABMs in serious gaming can help understand more complex interactions
Agent-Based Games for Testing Leadership and Management
ABM games with autonomous AI population
Test leadership style effectiveness
Explore which styles work best in different situations
Determine the best choice for a given scenario
Current state of the art is more conceptual
Advances needed in interfaces
Need to allow users to interact with simulation
Keep players engaged
Behavior Impacted by Multiple Factors
How different factors influence one another and result in behavior (opportunity consumption),
which aggregates over all simulated consumers and results in macrolevel outcomes that set the
conditions for a next behavioral cycle. In the consumat approach, the agents have existence needs
(e.g., food, income), social needs (group belongingness and status), and identity needs (personal
preferences, taste). To select a behavior an agent can employ four different types of decisional
strategies, depending on its satisfaction and uncertainty.A satisfied and certain agent will repeat its
previous demand, which captures habitual behavior/routine maintenance. A satisfied but uncertain
agent will imitate the demand of a similar other in its network, which reflects normative compliance
(fashion). A dissatisfied and certain agent will evaluate all possible demands and select the
one providing the best outcomes (optimizing). And finally, a dissatisfied and uncertain agent will
inquire the demands other agents had and copy this de ...
ITS 832Chapter 13Management of Complex Systems Toward Age.docx
1. ITS 832
Chapter 13
Management of Complex Systems: Toward Agent-Based
Gaming for Policy
Information Technology in a Global Economy
Professor Michael Solomon
1
Introduction
Simulating/Managing Social Complex Phenomena
Leadership and Management in Complex Systems
Serious Gaming
Agent-Based Games for Testing Leadership and Management
Single and Multiplayer Settings
Summary and conclusions
Simulating and Managing Social Complex Phenomena
Study of how people interact
Scale prohibits experimentation with real populations
Agent-Base modeling (ABM)
Networked agents
Each agent is an individual
Interaction may modify agent behavior
Managing complex phenomena introduces complexity
Techniques to manage turbulent situations vary
Technique success depends on responding to agent behavior
Which may change based on interactions
2. Leadership and Management in Complex Systems
Traditional leadership research
Generally focuses on single period in time
Doesn’t address dynamic relationships
Timing of leadership principle application matters
Primary leadership functions
Instructional and regulatory
Developmental
Simulations offer promise to help model leadership in complex
systems
Serious Gaming
Applying gaming techniques to real life situations
Flight simulators
Effective for evaluating complex environments
Player must interact with multiple actors and situations
Currently used for side range of training applications
Leadership use
Deterministic – limited scope
ABMs in serious gaming can help understand more complex
interactions
Agent-Based Games for Testing Leadership and Management
ABM games with autonomous AI population
Test leadership style effectiveness
Explore which styles work best in different situations
Determine the best choice for a given scenario
Current state of the art is more conceptual
Advances needed in interfaces
Need to allow users to interact with simulation
Keep players engaged
3. Behavior Impacted by Multiple Factors
How different factors influence one another and result in
behavior (opportunity consumption),
which aggregates over all simulated consumers and results in
macrolevel outcomes that set the
conditions for a next behavioral cycle. In the consumat
approach, the agents have existence needs
(e.g., food, income), social needs (group belongingness and
status), and identity needs (personal
preferences, taste). To select a behavior an agent can employ
four different types of decisional
strategies, depending on its satisfaction and uncertainty.A
satisfied and certain agent will repeat its
previous demand, which captures habitual behavior/routine
maintenance. A satisfied but uncertain
agent will imitate the demand of a similar other in its network,
which reflects normative compliance
(fashion). A dissatisfied and certain agent will evaluate all
possible demands and select the
one providing the best outcomes (optimizing). And finally, a
dissatisfied and uncertain agent will
inquire the demands other agents had and copy this demand if
the outcomes are expected to be
better (social learning).
7
Single and Multiplayer Games
AI may react poorly to management input
Simulating unexpected consequences of decisions
4. Overactive AI may degrade realism
Players can dynamically see how decisions affect others
Early simulations allow for only single players
Multiple real players adds more realistic interaction
Players replace some AI
Players interact with each other and AI
8
Summary and Conclusions
ABM-based gaming can measure behaviors of players
Supports experimentation in controlled environment
Study leaderships and management in complex systems
Focus
Interaction with leadership
Interaction with players as a result of leadership action
ITS 832
Chapter 15
Visual Decision Support for Policy Making: Advancing Policy
Analysis with Visualization
Information Technology in a Global Economy
Professor Michael Solomon
1
Introduction
5. Background
Approach
Case Studies
Optimization
Social Simulation
Urban Planning
Conclusion
Background
Assessing policy options for societal problems is difficult
Decision making methods
Data driven
Model driven
Visual decision supports helps in evaluating model output
Information visualization and visual analytics
Makes complex results accessible to many
Policy analysis
Part of process aimed at solving societal problems
Data Visualization
A characterization of data visualization by Stephen Few (2009).
Distinction between
activities addressed by data visualization. Exploration and
sensemaking as analysis tasks. Communication
as knowledge transfer task. Technologies differ with respect to
presented data (information
visualization for abstract data and scientific visualization for
physically based data) and interaction
capabilities of the technologies, e.g., graphical presentation
6. does not imply user interaction. Understanding
of provided information as intermediate goal. Good decisions
based on derived knowledge
as the end goal
4
Policy cycle
Policy cycle adapted from Jones (1970) and Anderson (1975).
Policy analysis is mainly
conducted in the policy formulation and the policy adoption
stage. While in the policy formulation
stage, alternative solutions to a given problem are defined; in
the policy adoption stage, one of
these solutions (policy options) is selected for implementation.
In a broader sense, policy analysis
can also be conducted in the other stages of the cycle. However,
in this work, we focus on the
pre-decision phase before a policy is implemented
5
Approach
Characterization of stakeholders
Policy makers
Policy analysts
Modeling experts
Domain experts
Public stakeholders
Bridging knowledge gaps
With information visualization (IV)
Cohesive view of model representation
Visual Support for Policy Analysis
7. Visual support model for policy analysis. The upper part
denotes the policy analysis
process which is conducted in the policy formulation stage of
the policy cycle (cf. Fig. 15.3).
Visualization is introduced into the process in order to support
the analysis of policy options
7
Approach, cont’d.
Synergy effects of applying IV to policy analysis
Communication - facilitated
Complexity - reduced
Subjectivity - reduced
Validation - improved
Transparency and reproducibility of results - increased
Case Studies
Optimization
Optimization of regional energy plans considering impacts
Environmental
Economical
Social
Social Simulation
Simulation of the impact of different policy instruments on the
adoption of photovoltaic (PV) panels by homeowners
Urban planning
Integration of heterogenous data sources in planning activities
Summary of Case Studies
8. Conclusion
Current model output is often difficult to understand
Not accessible for non-specialists
Information visualization (IV)
Makes model output more accessible
This paper applies IV to policy analysis
Contributions
Defined collaborations
Identified hurdles
Defined interface methodology