Section: Application Domains
Social sciences
“Visualization has thus far had less impact on the social sciences than the physical sciences, in part because of a dearth of funding for such efforts, but it holds the promise of effecting similar transformations.” [19]
From our experience, we see social sciences as a fertile area from which ideas can emerge, and where visual analytics techniques and methods can be designed, tested and validated. Because social sciences deal with non deterministic phenomenon, it places us right in front of challenges underlined by Thomas and Cook: to deal with large volume and inhomogeneous data, with constant changes in data, possibly making it ambiguous and uncertain. It is part of our agenda to develop closer relationships with research teams or industrial partners in social sciences.
We have had the opportunity to work in close collaboration with experts from quantitative geography through the ANR SPANGEO (Masses de données 2005 call) project(SPANGEO is a working group part of the S4 European initiative, see the URL http://s4.parisgeo.cnrs.fr/spangeo/spangeo.htm ). This ANR project has established close ties between individuals and long-term collaborations with the community research in geography. It gave us the opportunity to enter social sciences and explore the potentialities of interactive graph visualization and graph hierarchies for geographers [21] [38] , [67] , [36] , [80] [79] , [44] , [68] . Our approach clearly appears as complementary to classical cartography.
Cartographers and geographers, because they often stick to the usual geographical world map to depict statistical data, are limited by the size of the dataset they can visualize and thus visually analyze. Graph visualization offers them the possibility of visualizing and navigating whole datasets, at the price of leaving aside geographical constraints.
Moreover, quantitative geography also offered us the occasion to compare graph combinatorics with tools and approaches based on graph theory developed by geographers. The theory of small world network as initiated by Watts and Strogatz [90] , [92] , [91] draws new insights on spatial analysis as well as to systems theory. Its concepts and methods are particularly relevant to geography where spatial interactions are mainstream, and where interactions can be described and studied using large volume of exchanges or similarities matrix. In terms of geographical analysis of spatial networks, our methodology helped expert identify network entities acting as bridges between several components and offer a higher capacity for urban communities to benefit from opportunities and create future synergies.
Multiscale models . Our methodology exploiting hierarchical graphs [26] [7] [71] appears as a fruitful strategy to discover scales in datasets [21] .
Identifying structural changes . The data we study with quantitative geographers typically is time-stamped. That is, we often have data on populations, companies, air traffic, etc., collected through public surveys or by private companies over several years or months. The issues we now address is to identify structural changes or evolving patterns in networks. The task here is of great interest: the answer does not solely rely on mathematics or algorithms, but requires that experts link the identified pattern to real-life phenomenon and assess of its existence based on factual arguments (territorial policies, partnerships between companies, etc.).
Other partnership have been established with partners aiming at the development of visual approaches to hypothesis building and validation for law experts. The heterogeneous nature of the data will once again challenge us since law experts usually build cases based on information from texts (mail exchanges, newspaper extracts, etc.) and informal information from interviews or personal diaries.
