Accurate forecasts depend critically upon the ability to build behavioral models of the people and groups involved. Social scientists have traditionally constructed cross-cultural models by conducting either in-person or written surveys (1), or living with such groups (2), and then hypothesizing and testing correlations in collected data by means of various statistical models (3). None of these strategies will work in countries riddled with conflict like Iraq and Sudan today. Old surveys are likely to be outdated. Questionnaires and survey respondents may be influenced by the climate under which the survey is taken. In conflict situations, data must be gathered with real-time methods. However, building behavioral models in real time is particularly difficult (see the figure).

Computational social models may offer the best solution in cases where conventional data gathering is not possible. Tools such as The Resource Description Framework Extractor (T-REX) (4) use socio-cultural-political-economic-religious (SCPER) variables provided by social scientists in conjunction with other data sources (e.g., surveys), if available, and automatically extract relevant data from news sources, blogs, newsgroups, and wikis (i.e., collaboratively written information sharing sites). Other efforts such as the KEDS project (5) extract variables from specific news sources. The SCPER variables can include financial activities, violent event information, or political relationships. The source data can be automatically analyzed to recognize spikes in such activities, providing "early warnings" of potential conflicts. Unlike past methods, these methods do not require previous knowledge of the groups being investigated.