You can hardly tell where the computer models finish and the dinosaurs begin." Laura Dern, on the film Jurassic Park During the past quarter century there have been many developments in scientific models and computer codes to help predict the ongoing consequences in the aftermath of many types of emergency: e.g. storms and flooding, chemical and nuclear accident, epidemics such as SARS and terrorist attack. Some of these models relate to the immediate events and can help in managing the emergency; others predict longer term impacts and thus can help shape the strategy for the return to normality. But there are many pitfalls in the way of using these models effectively. Firstly, nonscientists and, sadly, many scientists believe in the models' predictions too much. The inherent uncertainties in the models are underestimated; sometimes almost unacknowledged. This means that initial strategies may later need to be revised in ways that unsettle the public, losing their trust in the emergency management process. Secondly, the output from these models form an extremely valuable input to the decision making process; but only one such input. There is a need to draw on much tacit knowledge which by definition cannot reside in a decision support system. Most emergencies are events that have huge social and economic impacts alongside the health and environmental consequences. While we can model the latter passably well, we are not so good at modelling economic impacts and very poor at modelling social impacts. Our knowledge of them is tacit and they lie in the complex space of Snowden's Cynefin categorisation of decision contexts. Thus we draw upon recent thinking in both decision support and knowledge management systems to suggest that we need a more socio-technical approach to developing crisis response system; and, in particular, we explore how model predictions should be drawn into emergency management processes in more balanced ways than often has occurred in the past.

No abstract

In recent years, there have been growing demands for a more participative approach to societal decision making and a higher level of accountability on the part of politicians and decision makers. Concurrently, the development of the Internet has provided an infrastructure to achieve these ends through substantive e-democracy. e-Democracy systems have the potential to draw on developments in decision support systems (DSS), involving stakeholders and the public in societal decisions. We argue that the key need in developing DSS for e-democracy contexts is to provide interfaces which explain the decision analytic techniques to users with a wide range of abilities, skills and cultures, using as an example an early application of automated explanations within a nuclear emergency management DSS. By building natural language context specific, sometimes graphical explanations, to explain and document the judgements needed in the process, this seeks to make the information from the application's domain knowledge broadly available and understandable. While recognizing that this application is limited, we believe that more sophisticated automated explanation facilities will be essential in e-democracy systems and we discuss the issues to be faced in developing these.

It is widely recognised in the social and management sciences that the effective support of decision-making requires a multidisciplinary perspective. This trend is also clear in nuclear emergency management (EM). However, communication between disciplines is not easy to maintain in EM contexts when the decision makers (DMs) are likely to be highly stressed. Such circumstances can lead them to revert to the instinctive patterns of perception of their core disciplines, making communication between disciplines difficult and, perhaps, obscuring complex interactions that have not been rehearsed in practice exercises. This paper explores decision making in EM and the nature of the socio-technical issues that will arise, suggesting that despite the lessons of past accidents the research EM community is still not taking a broad enough view of what future incidents may entail.

No abstract