The heuristic power of modelling is that it tests different potential interpretations by confronting them with empirical observation in models. In geography, modelling makes it possible to test the driving function of interactions in the emergence and evolution of the spatial organisation of social phenomena. Well aware of the essential part to be played by models in interdisciplinary communication, the members of the PARIS team are very attentive to developments in practice in other disciplines, and they operate reasoned transfers of methods and concepts developed in the fields of physics and computing, and more widely in the sciences of complexity. The aim is to enrich knowledge about the functioning of city systems and metropolitan spaces. After developing SimPop, the first model to be formalised using multi-agent systems in geography, the team is now engaged in further conceptual and technological advances, developing efficient modelling platforms and diversifying applications. The objective of the present research is to construct methods for evaluating, validating and experimenting on these models. A further challenge is the identification of objects and relationships within the models, which entails a formalisation of the notions of the individual, settings, and more generally ontology.
The team forms a reference for tools for the analysis of spatial configurations. For instance, gravitational methods using a smoothing function according to potential have met with growing success (use by the European Environment Agency, the ESPON programme etc). Certain theoretical foundations of these methods have been further explored in order to analyse the relationship between forms and processes in cases where there is a change in scale. The question for research is the relationship between the social space and the geometric shape.
Evaluation, validation, experiment
Several models developed on the SimPop platform enable experimentation on various scales of time and space. In addition to systematic work on comparisons, the team are developing applications that specifically take account of several geographical levels (rather than just two as in the classic applications in multi-agent systems), and they are developing new methods of assessment and validation that are suited to the range of scales envisaged.
Individuals, settings, ontology
In the multi-agent models being developed, the task is to formalise the effects of setting and of processes that relate to the logics of action by private of public actors, social groups, or institutions, combined with environmental processes, on different time scales. Theoretical consideration of the various entities associated with these different processes, of their formal status in the models, and of relationships existing among them, leading on to the definition of ontologies, enables better integration of these multi-scalar and multi-actor effects into the models.
Social and geographical space in the potential models
The smoothing operation that simplifies the representation of geographical information by generating continuous spatial distributions can be broken down into a set of projections onto different spaces: the social space of mobile individuals in time-geography is for instance projected onto the political-administrative space of census data (territorial grid), and the “smoothing” function projects the data in spatially discontinuous manner onto a continuous geographical space, i.e. that of the “smoothed” map (for instance a map of population potential or income). The analysis of these successive filters enables the generation of hypotheses on the behaviour of the actors.