Predictive Knowledge and Decision Making Process for Protecting Civilian Populations in Emergency Situations

Experts advising governments in various areas have long been using modeling techniques and simulations. Prediction technologies are fast increasing the capacity to simulate, visualize and anticipate disasters. In this sense, they represent technical innovations in governance and public action because they hold the promise to be able to anticipate and act on uncertain situations. However, the impact of these prediction technologies on decision-making in emergency situations has so far been little analyzed.

Through the presentation of the case study of the "Mercaptan cloud", we analyze to what extent the use of prediction technologies is likely to affect public decision-making processes for the safety of civilian populations. On the January 21^st 2013, a leak of Mercaptan gas from a chemical plant occurred for several days in France. Without being harmful to health, these gaseous emissions (smelling as the domectic gas) could be smelled on a very large perimeter, causing considerable inconvenience to the population. Emergency services, local and federal government and various institutional experts constantly interacted during the event to assess the danger.

From this case study prediction technologies that were used, will be reviewed. Their shortcomings due to socio-organizational constraints will be presented. These include: structural, institutional and individual mechanisms that shape the process of technical expertise and limit the direct translation of results into recommendations for deciding whether or not to evacuate the population. Meanwhile, different and complex uses of the technical expertise, broader than just the issue of the evacuation, were identified.

Although predictive technologies do not reverse the decision-making process, it will be highlighted that these predictive technologies can make it possible to propose alternative solutions to evacuation, and thus broaden the spectrum of public action. At least, this paper identifies social sciences approaches that could render the use of predictive technologies more accurate and realistic.