Concepts for society meta-modelling.

Zafiu, Adrian ; Ionescu, Valeriu ; Dascalu, Monica 等

1. INTRODUCTION

Simulating the human society is a difficult task as there are many factors that contribute to its evolution, but have an enormous potential (Epstein & Axtell, 1996). Previous simulations have centered on simple societies and particular economic and cultural aspects (Traulsena et al., 2009).

The complexity of a society is hard to model at macro level; therefore the solution chosen for this implementation is to describe the atomic components of the society's members and the environment that they interact with. If the model is well designed, it is possible to determine: the modifications necessary to be performed on the member's rules that will determine a desired evolution of the society; the avoidance of undesirable evolutions determined by the application of erroneous rules (Epstein & Axtell, 1996; Sullins, 2005).

2. ARCHITECTURE META-MODELING

2.1 Introducing entities, actions and events

The system is formed from a heterogonous environment (the passive element) where members evolve (the active element). The members are represented by intelligent agents that use in the simulation genetic algorithms and fuzzy rules. The genetic algorithms ensure the diversity of the population and the fuzzy rules describe the behavioral component and impose the microeconomic restrictions--representing the legislative environment where the evolution takes place.

The general data structure should ensure simple description of phenomena from the real world (Gilbert & Troitzsch, 2005); therefore we have analyzed the social, economical and political phenomena from real life and strategy game engines that refer to evolutionary societies.

The entities necessary to the simulator determined in the analysis stage are: agents, resources and groups.

The agent is the atomic element that consumes the resources having the purpose of living and moving in the simulated environment. The agents can be grouped with the purpose of better fulfilling their needs and faster evolving. The evolution of entities' attributes is ensured by genetic algorithms and is adjusted with fuzzy rules specific to the actions being performed.

Actions are reactions to events. Events are triggered when certain local or global states/stages are reached for the simulated system. An event is triggered by the encounter of two agents and when an agent discovers an accessible resource. Resources represent the elements required by the agents to live and to maintain the current living standards. Resources can be natural (when found in the explored environment) or artificial (when produced by agents).

[FIGURE 1 OMITTED]

Natural resources regeneratate depending on their type and artificial resources have a certain production speed. Production deficit and excess determine the necessity of product exchange between agent groups or their listing at the stock exchange. Direct trade between agents is based on direct negotiation and the stock exchange is based on an auction. Large quantities are always traded through the stock exchange.

2.2 Agents and groups

A man is a social being; therefore an agent can only exist in a group. Individual agents are represented by single agent groups. An agent can be part of multiple groups and a group can contain multiple agents simultaneously. Managing the multitude of agents and groups implies frequent updates on the relations between agents and groups. In order to reduce the processing time rare matrix were used, where the membership groups are stored in double linked lists on both horizontal and vertical. These have the advantage that the update takes a processing time of: O(1). The search operation is not necessary as it was incorporated in the evolution process of the simulators' entities.

The agent's attributes are: its level, the genetic inheritance, the skill and the group participation quota. The agent's level is determined by the accessed resource types and also determines the resources needed in order to maintain this level. The skill is the current capability to adapt to the performed task. The genetic inheritance influences the adaptation degree of the agent to different tasks that are performed. The group participation quota is an abstract value (equivalent to money) and represents the individual's intake to the group as a result of the actions performed. An agent does not hold resources. The agent's sex is an arbitrary value established at the creation time. Each agent lives a certain number of cycles, with maximum and minimum values, depending on the genetic inheritance and its skill.

[FIGURE 2 OMITTED]

The group's attributes are formed by resources that it can make available to its members, proportional to their quota. Each contribution from an agent updates its respective quota and reduces the other agents' quota. The resource combination in a group determines the level of that group. The evolution of the group's individual members towards the group level is based on their participation. The relationship between the number of agents in a group, the group's resources and the group's level (and its members) determines the wealth degree: poor, average, rich. Only the rich members (as in the agent level) can go to a superior level. The cost of a superior level of the agents is given by the quantity of resources necessary in each cycle. The wealth level is determined by the number of cycles in which an agent can exist based on its participation to different groups.

The agent's affiliation to a group is determined by the group's needs and the agent's abilities. The need is given by the group's deficit of a certain resource and determines its wish to include a member with a high ability in collecting the resources that are in deficit. Each resource of a group can be catalogued as: insufficient, deficit, sufficient, surplus, and excess.

The necessity to eliminate an agent from a group is apparent when its abilities do not allow maintaining its level in the group. An agent can also leave the group when it cannot ensure its minimum level of resources from the contribution brought to the group.

2.3 Resources

The resources collected can be natural or artificial. Natural resources have a certain regeneration specific to the resource type and the characteristics of the environment they are in. Artificial resources are characterized by the production speed and the characteristics of the environment they are being produced.

We consider the ability to collect resources equal to the ability of the agent to produce resources. Creating artificial (production) resources is a result of a group decision. Maintaining the production level is monitored by a special agent that also handles their collection. The other agents in the group can help collect the resource. Agents in other groups do not have direct access to the production resource. Creating a production resource implies a cost recovered in time through the intake from the produced resources.

Between resources we propose a dependency relationship. A resource can be collected or produces if there are available other resources and, sometimes, by consuming other resources. The purpose of the simulation is modeling the real world; therefore building a good dependecy tree is the key to the whole application (Fig. 4).

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

[FIGURE 6 OMITTED]

3. CONCLUSION

The concepts described above allow the creation of a computer simulation of a society (Dascalu et al., 2008). The proposed concepts and relationships were implemented in an early stage of the simulation with 400,000 agents and 5000x5000 landscape. The resulted simulation step is seen in figure 6.

Attribute tuning will be necessary to determine a stable social evolution. Further research will be necessary in order to incorporate other concepts that have a deep impact in the society structure (such as family an an entity between individual and group, etc.).

4. ACKNOWLEDGEMENTS

The work for this paper was done with the financial support from 731/2009 project from Romanian IDEI research program.

5. REFERENCES

Dascalu, M.; Franti, E. & Stefan, G. (2008). Artificial societies, a new paradigm for complex systems' modeling, Proceedings of the 7th WSEAS International Conference on Artificial intelligence, knowledge engineering and data bases, Cambridge, UK, ISBN:1790-5109

Epstein, J.M. & Axtell, R. (1996). Growing artificial societies social science from the bottom up, ISBN-10: 0-262-550253, ISBN-13:978-0-262-55025-3

Gilbert G. N. & Troitzsch K. G. (2005). Simulation for the social scientist, New York, ISBN -13 978 0335 21600 0 (pb)

Sullins J. P. (2005). Ethics and Artificial life: From Modeling to Moral Agents, Kluwer Academic Publishers, vol 7, Pg. 139--148, ISSN:1388-1957

Traulsena A., Hauertb C., De Silvad H., Nowakb M.A. & Sigmunde K. (2009). Exploration dynamics in evolutionary games, Princeton University, Proceedings National Academy of Sciences of the USA; 106(3): 709-712