Context of the SWIFT project
The state of Victoria in Autralia frequently faces more or less serious bushfires. Despite information and awareness campaigns, the population does not always adopt the best behaviour in response to bushfires. As a result, many victims are people waiting too long before evacuating the risk area. Reports and sur- veys show a discrepancy between people’s perception and the actual situation. On the one hand, the population is not really aware of the possible actions they can take, and is therefore often too passive. On the other hand, the risk managers expect people to behave ”rationally” but are surprised by their ac- tual behaviour; besides the messages they send are often not adapted to the population’s expectations. As a result such messages might be discarded as not relevant by the targeted population. Reports have shown that when looking for information about the bushfires, people can turn to social networks, despite not completely trusting them, but because they are faster and provide a sense of community, and because they might ignore the official information channels.
The SWIFT project is interested in solving this mismatch by using a serious game. This serious game can be used with the population to raise their aware- ness of best behaviours in case of a bushfire. But it can also be used by risk managers to test different strategies in informing and warning the population, and to better understand their reactions.
Goal of this M2R project: a model of population trust
The focus of this M2R project is on a serious game for emergency managers to test various strategies for informing and alerting the population. In order for this serious game to be helpful, and in particular to lay valid results (a strategy that works in the game should work for real), the model of the population should be as accurate as possible. The objective of the project is to formalise and implement such a model, with a focus on how people develop and maintain trust in information sources. Interesting questions to be explored include:
Criteria to choose and determine trust in a source of information: speed of delivery, availability, reputation, etc
Influence of emotions of the population on their trust (do some emotions make people more receptive?)
Influence of emotions of the sender (does a person expressing specific emo- tions look more trustworthy?)
Influence of personality and profiles of behaviour on trust
The project will start with a review of relevant bibliography about trust, emotions, and behaviour in Australian bushfires.
Evaluation: prototype
The model of population behaviour developed above will then be implemented in the GAMA multi-agent simulator. A first prototype of the serious game will be developed, allowing to compare a few information strategies and validate the model. Possible strategies to be tested:
Quantity of information: redundancy vs cognitive overload
Use a trusted intermediate agent (community leader) to forward the information to the population in that community
Appropriate time of delivering the information
Emotional charge of the message
The first results obtained with this prototype might lead to a publication.
Short bibliography
Carole Adam, Elise Beck, Julie Dugdale. Modelling the tactical behaviour of the Australian population in a bushfire. ISCRAM-med 2015, Tunisia, October 2015. Springer LNBIP.
Carole Adam, Elise Beck, Julie Dugdale. SWIFT: simulations with in- telligence for fire training. Poster. In Proceedings of ISCRAM 2015. L. Palen, M. Buscher, T. Comes, and A. Hughes, eds.
Why don’t they do what we think they should - understanding people’s response to natural hazards. By Alan Rhodes of Emergency Management Victoria.
Forgas, J.P. (1995). ”Mood and judgment: The Affect Infusion Model (AIM)”. Psychological Bulletin 117 (1): 3966.