Wednesday, 22 July 2009

Literature Review Agent-Based Simulation "On generating hyphoteses using computer simulation"

Conclusions that i get after I read article entitled "On generating hypotheses using computer simulation" by Kathleen M. Carley that was published in the Proceeding of the 1999 International symposium on commend and control research and technology.

What is the use of computer simulation?
  • Computer simulation can be used to develop the theory and generating hypotheses

Why computer simulation is important to generate a hypothesis?
  • Because, social interactions are dynamic, adaptive and non linear.
  • As an example of non-linearity is, the declining ability of an information to shift agent’s perception.
  • As the number of information that support an idea believed by an agent increase, the impact of dis-confirming information decrease.
  • Such non linearity will bring difficulties for the researchers to infer the impact of a learning process, adaptation and agent's response, especially in a dynamic environment.
  • Computer simulation can help researchers to think about the impact of such non linearity and to generate a number of consistent hypotheses.
  • To generate hypotheses using agent based simulation, researchers can conduct virtual experiment.
  • Obtained hypotheses then can be tested through experiment in the real world.

Steps to conduct good virtual experiment
  • Identify core variables: Core variables are parameters that are assumed to be the most relevant variables affecting the dependent.
  • Define the range of parameters that are going to be explored.
  • Set the non-core variables as random numbers or according to the real world data.
  • Run the simulation multiple times for each experiment set. Ideally researchers could obtain much larger data than, that can be obtained in real world experiment.
  • Analyze the result of simulation statistically.
What are the advantages of agent based simulation compare to the human laboratory experiment?
  • Using agent based simulation, researcher can explore range of parameters and type process that are impossible to be explored in the real world.
  • These impossibilities can be caused by the high amount of cost or by the ethical issues.

Monday, 20 July 2009

Literature Review Agent-Based Simulation "What is the truth of Simulation?"

Conclusions that i get after I read article entitled "What is the truth of simulation?" by Alex Schmid that was published in Journal of Artificial Societies and Social Simulation Vol.8, no. 4 (2005)

What is validation?

Validation is a process to determine the sufficient accuracy to which a model or simulation is a representation of the real world system from the perspective of its specific purpose.

There are two key components of validation process:
  • Model’s accuracy that is measured with scale from 0 to 100%
  • Model’s validity mean whether a simulation model is true or not to in the perspective of its specific purpose.
In order to deeply discuss the validation concept of a simulation model according to the scientific framework, it would be appropriate to first discuss the philosophical concepts of truth.

Are simulation truths worthy?
What is truth worthy?
  • Philosophically an object classified as truth worthy object if that object can be judged as true or false.
  • Traditionally, linguistic object such as statements and judgment are the main objects of truth.
  • If we can accept mathematics as foundation of simulation as a language, and the communicative aspects of a simulation of the model then, simulation model it is a truth worthy object.

There are some appropriate criteria of truth that can be applied to a simulation model.

1) Correspondence theory of truth:
  • A statement is considered true when it has a correspondence with a fact in the real world.
  • A simulation can be considered as true if its components refer to the facts in the real world.
  • Correspondence theory of truth can be useful in judging the accuracy of simulation model.

2) Consensus theory of truth:
  • A simulation considered as true according to the consensus theory of truth if it can be rationally accepted in an ideal or optimal condition.
  • Whether a simulation model is correct or wrong will depend on whether the structure of the simulation can be accepted or not by others who think rationally.
  • Consensus theory of truth can be useful in judging the validity of simulation model.
3) Coherence theory of truth:
  • A simulation model can be considered as true according to the Coherence theory of truth if its components have a root to the branch scientific disciplines that is believed to be true.
  • Coherence theory of truth can be useful in judging the validity of simulation model.

Thursday, 9 July 2009

Literature Review Agent Based Simulation "Why Model?"

Summary results of the work of Joshua M. Epstein article entitled "Why Model?" that was published in Journal of Artificial Societies and Social Simulation 11, no. 4 (2008)

  • Basically, any researcher who tries to make a projection or imagining a social dynamics is running a model.
  • The most important thing for a researcher is, whether he is able to make an explicit model or not?.
  • In an explicit model, all the assumptions used are clearly specified, so that their impact can be tested.
  • By creating an explicit model, we can combine expertise from various fields such as Biomedical and ethnographic.
  • Parameters in an explicit model that can be calibrated with the historical data and, its behavior can be tested with the present data.
  • In an explicit model, sensitivity analysis can be done by, sweeping huge range of parameters over vast range of possible scenarios.
  • By running a sensitivity analysis, researchers will be able to identify uncertainty, region of robustness, and important threshold.
  • A model should not always able to predict.
  • However, the ability of a model to uncover trade off, sensitivity and uncertainty can guide the decision making process.
1) Explanation
  • For example, the electrostatic model can explain how a lightning occur, however, it cannot predict when and where the lightning will appear.
  • A simulation model can explain the emergence of a pattern caused by interactions of a number of agents.
  • This kind of explanation is called generative explanation.
2) To guide data collection
  • Many researchers have been wrongly applied the inductive method by gathering as many data as possible then run regression on it.
  • A model is used only as a calculation tool of data.
  • In fact, to process data with a model, a number of assumptions must be fulfilled.
  • For example, the existence of radio waves were first detected through the Maxwell equation, after that the supporting data observed.
  • Without able to specify the assumptions used in a model, researchers are not always clear which type of data should be collected.
3) Illuminate core dynamics
  • A model can be used to clarify an abstraction, and strengthen human basic intuition .
  • Thus, although the model contains a simplification, the model still can be useful.
4) Suggest Analogies
  • A number of varieties of processes that does not seem related can have the same formal form.
  • For example, the algebraic form of Coulomb law is identical to the Newton's law of gravity.
  • By creating a model, we can make an analogy of a process by the other process and compare the behavior of both processes.
  • If the behavior of analogical model is similar to the behavior of the target process then, there is possibility that laws and theories in the analogical can also be applied the target process.
5) Rise new question: models can surprise us and increase our curiosity, and lead to new question.

Friday, 3 July 2009

Literature Review Agent Based Simulation "The Bigger Picture"

Summary results of the work of Tamas Viscek article entitled "The Bigger Picture" that was published in Nature Vol.418 11 July 2002 pp: 131

  • A complex system is a system that, its overall behavior requires different theoretical explanations from the theoretical explanations that are used to explain the behavior of its sub-system
  • Each level in the complex system, regulated by different laws
  • Both deterministic and random features, owned simultaneously by a complex system
  • A complex system has chaotic behaviors. The system can show a regular behavior however, may change dramatically and stochastically in space and time due to small changes in the initial condition
  • Because, the universe is consisting of many components at various levels, these components are connected and interact with each other.
  • Interaction between components that occur at each level, produce a behavior that requires different interpretation of the results of interaction on the other levels.
  • Behavior of the whole system, emerge as a result of the interaction among components in same the level and among components on different levels
  • Complexity science is a field that studied the process of behaviors formation of a system.

  • Traditionally, human try to understand the nature using reductionism perspective (through simplification and analysis)
  • Reductionism has a weakness because, ignoring a number of factors that works simultaneously, if their impacts are considered not significant
  • Complexity science considers all processes that occur simultaneously on different levels important
  • Behavior of the system as a whole will depend on the results of these processes in a non-trivial way
  • By creating a model of a system researchers can understand and manipulate the behavior of the system as a whole.
  • Computer simulation is a tool that can improve our insight about the mechanisms that occur in a complex system.