A Modeling and Simulation Framework for the Social Sciences

1. Motivation

• The variety and complexity of social processes could be modeled more appropriately by supplying an integrated view of different modeling approaches.
• Providing a toolkit which especially does not bother a user with programming and implementation details would be useful for teaching social science students in modeling and simulation tasks.
• It would facilitate the communication and exchange about simulation models within the scientific community (e.g. transparency of models, reuse of model components).

2. Requirements

The simulation life cycle

• model description: provides a modeling formalism executable by software systems
• experimental frame: includes several phases, which have to be executed by a researcher for model verification, validation and after all the application (e.g. for explanation, forecasting) of simulation models, in an iterative way:
  • Model initialization (including empirical data input)
    • Databases for input
    • Random number generation
  • Simulation runs
    • Experiment description
    • Control of simulation parameters (e.g. time intervals, simulation time)
    • Nonnumerical visualization (e.g. animation)
    • Logging and debugging features (e.g. run time tracing, dumping of input and output)
  • Model analysis
    • Nonnumerical visualization
    • Statistical data analysis
    • Sensitivity analysis
    • Parameter optimization
  • Presentation of results
    • Graphics
    • Report generator

Kernel features

• Integration of the simulation life cycle
  

  

  Apart from the model description phase, which has to be the starting point of each simulation, the other phases described above will not be executed in a strictly prescribed order. Usually, experimenting with models requires the iterative and multiple application of model initializations (e.g. empirical data input), simulation runs, model analysis methods, and presentation of results. Combining these time consuming tasks in a comfortable and transparent way is an essential requirement of a simulation system.
• Integration of different modeling techniques/hybrid models
  In the last forty years a lot of approaches have been used to model and simulate human social behavior, which were mainly restricted to certain modeling aspects. Additionally, the usage of different approaches by different researchers (implemented in different simulation systems) has prevented substantial discussions about simulation models within the scientific community. The integration of these approaches requires the answer to some questions beforehand: What are the characteristics of common social science modelling approaches? Which "new" approach or paradigm can be found (or can be developed) that incorporates all the earlier approaches and helps to formulate new problems (and, eventually, helps to solve them) as well. Actually, we think that multi-agent modeling based on concepts from Distributed Artificial Intelligence (DAI) could be a solution. Generally speaking, DAI is concerned with sets of units (agents) existing in an environment which they have in common. They communicate among each other and cooperate in the solution of complex problems by adding their particular capabilities.
• Integration of different user/usage levels
  Simulation is applied usually by different user groups with different abilities and purposes:
  • Simulation and programming experts: write their own simulation programs from scratch by using programming languages
  • Social science researcher, simulation tool developers: combine basic modules into simulation models for new research or to standardize simulation models
  • Teachers: put together models of common modeling approaches for teaching social science students in formal modeling and simulation techniques
  • End users, lay persons, politicians, decision makers: apply simulation models which are represented graphically by parameter variation
  The concept of a more general modeling and simulation toolkit for the social sciences should not prescribe a specific user level. Instead, a user should have the option to choose an appropriate level and to change it as well. This requires the realization and integration of different user/usage levels, stacked on top of each other and transferable between each other.
• Abstraction/Refinement mechanisms
  According to concrete modeling and simulation purposes the user must be able to choose an appropriate model granularity. This helps to avoid unnecessary model details and at the same time supports step-by-step refinement of model components.
• Platform independence
  The toolkit should run on different computer (operating systems) systems (Windows 3.1/95/NT, Unix, Solaris, OpenStep, ...)

Extensions

• Multilingual support
• Distributed simulation (phase/model oriented)
• WWW version
• Explanation feature

3. Architecture

Integration of different modeling techniques

• modeling agents: which structure do agents have, in which way will they be able to act / react, which capabilities will they need etc.? For the moment, we keep the distinction of three kinds of agents:
  • Reactive agents: react to messages from their surroundings by sending other messages to other agents and by actualizing the inner representation of their surroundings. All this happens according to fixed rules or plans which cannot be changed by these agents.
  • Intentional agents: have the same capabilities as reactive agents. Applying ``meta rules'', they are moreover capable of defining goals, e.g. depending on their motivation or their needs. They can detect conflicts between goals, set priorities, and design plans to achieve their goals, and they can be informed about each other's goals, assumptions, and actions.
  • Social agents: additionally have explicit models of other agents. This is why they are capable to reason about other agents' goals, expectations, motives, and capabilities, and to include them into their action plans.
  Because of the fact, that social systems are structured hierarchically (e.g. families, households, groups, societies), social science models require a further distinction in:
  • indivisible agents: do not consist of other agents
  • aggregate (or systemic) agents: made up of other agents, but may interact with their surroundings in the same manner as indivisible agents
• constructing societies: in which way will agents interact, in which manner is interaction controlled in order to solve a problem, which kinds of organization are needed to facilitate coordination and/or cooperation etc.?

Integration of different user/usage levels

• the programming level: programming simulation models in a general purpose object-oriented programming language, using interfaces to existing agent definitions,
• the module level: combining agent descriptions and interaction structures into simulation models,
• the scheme level: defining simulation models for classes of models, based on scheme models,
• the application level: developing model class specific simulators and at the same time making problem oriented user interfaces available.

4. Plan

• Discussion of this paper
• Collaborative development of detailed requirements and software architecture
• Consideration of existing products/projects (market research)
• Funding

5. Participants

(until now)

• Kai H. Brassel (Darmstadt, D)
• Edmund Chattoe (Surrey, GB)
• Ottmar Edenhofer (Darmstadt, D)
• Michael Möhring (Koblenz, D)
• Claudia Pahl (Duebendorf, CH)
• Elke Schumacher (Koblenz, D)
• Klaus G. Troitzsch (Koblenz, D)

(tentative)

• iec PROGAMMA (Groningen, NL)

University

Computer Science Department

WWW