A core element in agent-based social simulation research is the analysis of emergent processes or phenomena. Identifying and understanding such processes supports the identification of both opportunities and dangers. Participatory agent-based social simulation enhances this identification by involving stakeholders in the design and evaluation of the simulation systems. While substantial work in social simulation is highly formal, participatory approaches are concerned entirely with the representation of observed actors, social processes and institutions.

The close connection between observation and modelling is an advantage over other modelling approaches such as economic modelling or system dynamics which rely on unrealistic assumptions, such as perfect or limited information. The issues of concern to the social simulation research community entail behaviour of actors and software agents in conditions where the amount of information far exceeds the information processing capacities of the actors or agents.

The compositional methodology used here entails the software representation of agents at different grains of analysis. The purpose of a coarse grain analysis is to reduce the number of entities being considered so that "manageable" numbers of relationships can be represented. Nonetheless, understanding reality at a coarse grain requires a corresponding analysis of behaviour at a finer grain with more detailed specifications of the components of the coarse grained models. Finding consistent models at different grains will facilitate the explication of emergent macro outcomes from micro behaviour and how these macro outcomes themselves constrain the micro behaviour. Modelling emergence is a common feature of ABSS research while the relationship between individuals and social structures has been well explored by sociologists both formally (Conte, Castelfranchi) and qualitatively (Giddens, Bell, Law).

Ensuring the model is a good representation of the real phenomena is the point of validation. It implies a homomorphic mapping from the data-extracted relations to the model characteristics. Compositional methodology demonstrate at this point that it also allows intermediate validation and helps keeping a closeness between observations and modelling of the underlying phenomena.

The model elaboration is describing in a formal way the phenomena we observe. Nevertheless, it is not reasonable to think that any perfect representation can exist in studies of social phenomena. This awareness leads to the implementation of random factors in order to represent hidden "side effects" or links. The means of implementing such side effects will be developed following Moss (1999). The validation process will involve stakeholder participation to identify politically sensitive issues as well as to assess the descriptive accuracy of the process representations.

This combination of stakeholder participation, ABSS and the compositional methodology has no precedent. Consequently, a key feature of the research will be to develop and test the new research design within the overall framework of the EU-funded FIRMA project.

The purpose of this project is to contribute to the development of a novel research design. So far, in developing policy relevant models, the description of the underlying relevant system, whether verbal or formal, was incorrect and incomplete. It is necessary, in order to use it as a tool, to have the possibility to query a model and understand in detail the emergence of results.

The agent based approach is used and implemented in SDML, and is incorporated in the core model of the FIRMA project, resulting in a modelling technology with required characteristics. (aim 1)

This agent representation will be designed based on models derived from genetic programming application and demographic data collection and treatment. Distinguishing types of agents and selecting the relevant ones should lead to this details understanding. (aim 2)

The use of different grains is an efficient way to keep simplicity and faithfulness. A coarse grain can be specifically used for policy analysis, whereas finer grain will be more interesting for domain experts or stakeholders interventions. The latter is also the main source of information about behaviouring details, or even patterns emergence. (aim 3)

The CPM already developed genetic programming techniques with a previous project (Integrated Marketing Intelligent System). It is used to impute water demand model from numerical data. But relevant component specifications of water demand are necessary. The available numerical data in the present case includes aggregate water use for residential districts, temperature, humidity, wind speed and direction, all provided by Thames Water PLC. Several plausible models will result, and the selection will be guided by the domain experts and the stakeholders involved in the FIRMA project with respect to statistical and qualitative evidence. (aim 4)

The nature of the participatory agent-based analysis implies a close relationship with decision-makers. Consequently the comments and information exchanges between modellers and stakeholders are more likely to happen under the shape of interviews more than surveys, even if the novelty of this approach implies some uncertainties about it.

Bell E., 1999, The negotiation of a working role in organisational ethnography, Int. J. Social Research Methodology, vol. 2, No 1, pp. 17-37.

Castelfranchi C., 1997, Challenges for agent-based social simulation. The theory of social functions, SimSoc 1997, Cortona, Italy.

Conte R. and al., 1997, Lectures notes in economics and mathematical systems: simulating social phenomena, Springer-Verlag, Berlin.

Giddens A., 1984, The constitution of society, Cambridge, England, Polity Press.

Law J., 1997, Traduction/trahison – Notes on ANT, WP, Department of sociology, Lancaster University, http://www.lancaster.ac.uk/sociology/stslaw2.html.

Moss S., 1999, Relevance, realism and rigour: a third way for social and economic research, CPM report no 99-56, http://www.cpm.mmu.ac.uk/cpmrep56.html.