IDEAS Factory - Game theory and adaptive networks for smart evacuations.

Lead Research Organisation: University of East London
Department Name: Cass School of Education & Communities


In the United Kingdom the evacuation of cities has not been a recent concern although incidents such as floods or acts of terrorism have meant that cities have had to be partially evacuated. Whether the area to be evacuated is a city district, a whole city or even a city region the way in which people behave and how they choose to share and receive information is paramount. In the pre and warning stages of an evacuation people will behave in different ways. Some individuals might have access to rumours that a crisis is about to occur (we call these weak signals) and start evacuating before the official order to do so. Others might have no access to information, may decide to act only when others do, or even ignore official warnings and remain in their homes (as was the case with Hurricane Katrina). During the evacuation there will also be communication between the authorities and emergency services and even between evacuees who might communicate through mobile telephones or social networks (such as twitter). Even at the end of the evacuation individuals will be making decisions about what to do (to stay in the evacuation zone or return to the city) and will stay in touch with others. In this project we consider that the ways in which individuals might behave in evacuations can be described using mathematical models and theories. One way of describing how people act as a collective is through agent based models, where individuals influence others information and behaviour, and another is game theory, where individual's decisions are strategic. By combining these we can predict how individuals might behave in an evacuation and determine the ways in which the emergency services and local responders might influence social networks and communication to help individuals and businesses evacuate safely and effectively. Of course, we anticipate that individuals might have access to new technologies in terms of communication so part of the project will look at how this may change the dynamics of evacuation.The project is informed throughout by current practice and will inform policy. Three cities (London, Birmingham and Carlisle) are chosen as case studies of how real evacuations might be assisted by the project. We will look at the evacuation plans in each of these cities, conduct focus groups and expert interviews with relevant parties, and analyse the use of social networks in actual evacuation or crisis events. Stakeholder groups will comment on the models which we produce and feedback as to how they can be adapted to work in practice.The results of the project will feed into how real evacuations work and inform the strategies and technologies used. We will produce both an interim and final report from the project and our impact plan will make sure that relevant parties are informed about the project and its results. One example of how the project might be useful is in terms of informing emergency planners and services in cities. If these planners know how social networks and mobile communications are used in an evacuation then they can tailor their messages accordingly. If, for example, there are two roads out of a city and analysis of social networks and traffic flow shows that individuals are intending to use the first road which would cause congestion then they can decide what sort of message would best direct people towards the second road, how long it would take to propagate and what effect it might have. In another example, road signs do not often give real time information about what is happening, which might be crucial during an evacuation. By using electronic and ambient signage not only can a message be sent but the effect of the message can be judged.

Planned Impact

The ultimate aim of this project is to contribute to improving the efficiency and the safety of city evacuations. If successful in its aim, this project has not only the potential to fundamentally change the way in which policy makers, public managers and businesses operate in the context of emergency planning and response, but it will also have a direct impact on the lives of people when an emergency situation occurs. The ultimate beneficiary of this research will, therefore, be the general public. In the case of an actual evacuation the outcomes of this research will translate into a reduced number of injuries and fatalities. More specifically, faster and safer evacuations will lead to a number of additional socio-economic benefits including a reduction in potential policing costs, fewer hospital admissions and robust business continuity management. If communication can be tailored to specific groups of the population, then there will be more opportunities for a coordinated collaboration with the general public and less chance of social conflict arising from emergency situations. We will make sure that the public benefits from this research through working with stakeholder groups and policy makers in translating the results of our project into practical outputs that lead to safer and faster evacuations. In the design of our project we have built in collaboration with, and dissemination to, stakeholders. In terms of public awareness the project team will maintain a high media profile by using established links with local, national and international media: press releases will be issued on a regular basis and feature articles will be prepared for the local/national press; radio panels and interviews with key stakeholders and experts in the field will also be facilitated. Economically, there are additional potential cost savings if evacuation can be made more efficient for emergency planners to administer. By linking to existing on-line social networks, exploiting mobile technologies and ambient computing, emergency planners and services will be able to collectively and collaboratively foster links with emerging and established real-life networks and tap into street-level knowledge in real-time. This will ultimately translate into a more efficient use of available resources and reduced cost in emergency planning and response. It will also feed into the government's CONTEST 2 strategy. This level of synergy between the general public, businesses, policy makers and emergency services will be achieved by actively engaging with representatives from across public, private, community groups and the voluntary sector. In order that these benefits to the commercial sector, savings to the general public and benefits to the economy might be maximised, we have created a work package on policy advice (WP4). This package is designed to maximise the impact of our findings on business strategy and policy with regard to continuity and resilience in evacuations. In terms of technological development, the project will actively seek further partnerships and target all potential beneficiaries through a number of events and activities including seminars. The purpose of such seminars is threefold: (1) to disseminate emerging research findings to all potential beneficiaries; (2) to seek future partnerships and further financial support (e.g. through a Knowledge Transfer Partnership); (3) to seek out potential consultancy opportunities and collaborations with industries. To capitalise on the above impacts a high profile London-based dissemination event will complement the above mentioned activities and it will coincide with the launch of the final report. The event will target 50 key stakeholders from UK organisations. Invited speakers will include members of our advisory board to allow knowledge transfer and expertise dissemination across all groups.


10 25 50
publication icon
Black AJ (2012) Mixing times in evolutionary game dynamics. in Physical review letters

publication icon
Bladon AJ (2011) Learning dynamics in public goods games. in Physical review. E, Statistical, nonlinear, and soft matter physics

publication icon
Galla T (2011) Optimizing evacuation flow in a two-channel exclusion process in Journal of Statistical Mechanics: Theory and Experiment

publication icon
Galla T (2012) Fluctuations in meta-population exclusion processes in Journal of Statistical Mechanics: Theory and Experiment

publication icon
Heaton LL (2012) Advection, diffusion, and delivery over a network. in Physical review. E, Statistical, nonlinear, and soft matter physics

publication icon
James King (Author) (2011) Discrete Optimization and Data Analysis

Description · Social media presents a challenge for city evacuation policy. It is most effective when information is accurate and timely, when the city has an orientation towards social media; when it is used in conjunction with other forms of media (in terms of increasing both the proportion of the population informed and equity) and when it can be used to source outlying information. However, if agents are 'over-informed' it may result in congestion and jamming; it has a selective demographic effect; it is less effective than old media as a mechanism for an initial warning.

· Different cities have different orientations to using social media in an evacuation and this should be considered in city wide social media strategies.

· Network coverage is an issue in some regions if cellular communications are to be systematically used in engaging with the public in an evacuation.

· Old media is still the most effective medium for warning and informing the general public. Social media may increase the effectiveness of warning and informing, but the impact is marginal.

· Interactions between 'old' and 'new' media are important in terms of not only the spread of information but (potentially) in terms of the ways in which content is relayed (transmedia).

· There is huge time sensitivity in terms of the effectiveness of warning and informing the general public. There remains the problem of warning the population in the middle of the night when traditional warning systems may be more effective (sirens, door to door).

· The issue of how opinions are formed on social networks is central in discovering how this may impact upon city evacuations. It is interesting to look at how prior beliefs evolve as individuals swap information on social networks (noting a duality between coupled urn models and models of Bayesian inference by a population) this becomes both a mathematically rich and relevant topic for further investigation

· For social media to be of use by citizens in evacuations information should ideally be both accurate and timely.

· In an evacuation it is plausible that social media communications will follow situational awareness categories over time (Perception, Comprehension, Projection) and that this has implications for engagement with social media by authorities.

· Communication between agents does not always help. In one of our evacuation simulations if the number of communicating agents exceeds the optimal ratio then agents are `over-informed' and frequently change their minds during the process, slowing down the overall efficiency of the evacuation.

· In a simulation of the evacuation of a UK city, communication between agents caused congestion.

· Using statistical methods we can identify outliers in social media in evacuations. This is useful both for identifying novel and troublesome social media outputs.

· For emergency response to be effective in a city evacuation algorithms used to allocate emergency personnel to events should be changed according to the nature of the situation

· Disaster education in an evacuation / invacuation should cover a wide pedagogical spectrum to take into account different types of learning.

· The use of social media for disaster education may be technologically progressive but is not always progressive in terms of equity and representation.
Exploitation Route The research has implications for the design of evacuations and social media planning for cities and governments. This research has been disseminated to local authorities, government departments and international organisations.
Sectors Communities and Social Services/Policy,Digital/Communication/Information Technologies (including Software),Education,Government, Democracy and Justice,Security and Diplomacy

Description The work on 'disaster education' has included an important element of co-production. At the level of UK policy, the advisory board of the EPSRC project comprised senior representatives from UK government (Cabinet Office, Home Office, Government Office for Science) who were involved in guiding the design of the research field. Furthermore, a requirements analysis was carried out involving emergency planners in three cities (Birmingham, London and Carlisle) in 2010 and the results of the project were reported back to them in 2012. This has resulted in several policy and practice impacts. Firstly, the emphasis of the project on social media in disaster education (discussed in Preston, 2012) has led to an increased emphasis on social media for disaster preparation in the three cities. Secondly, the Cabinet Office also consider that it is important that responders in each local authority have information about the various forms of pedagogies which can be used in disasters and have asked Professor Preston to produce guidance specifically on 'disaster education' to be distributed through their networks to local authorities, first responders and blue light services. This will be distributed nationally in early 2013. Although it is difficult to judge the exact level of significance in this area, the research has led to a reappraisal of forms of disaster education. As evidence of this, Preston and Kolokitha were asked by Essex Fire and Rescue Service to appraise their social media education strategy. This led to changes in the way in which they communicate disaster education to their constituency. At an international level, the Department of Homeland Security invited Preston to present his research at their 2011 Science Conference. This was the only UK project presented at the conference which helped the DHS in terms of their evidence in this area . The research has influenced cities and government in terms of the way in which they consider mass evacuations. Beneficiaries: Practitioners, 'blue light' services, government Contribution Method: By enabling local planners and government to use the results of inter-disciplinary social and physical sciences in planning for mass evacuations.
First Year Of Impact 2010
Sector Communities and Social Services/Policy,Education,Government, Democracy and Justice,Security and Diplomacy
Impact Types Societal

Description City resilience impacts
Geographic Reach Local/Municipal/Regional 
Policy Influence Type Influenced training of practitioners or researchers
Impact A series of focus groups and workshops were held with City Resilience forums in Birmingham, London and Carlisle as well as in Essex and Devon (Local Resilience Forums). These considered how best the results of the project could be implemented in terms of including policy and practice on evacuations.
Description Department of Homeland Security
Geographic Reach North America 
Policy Influence Type Citation in other policy documents
Impact The project was invited to present at the Department of Homeland Security Science Conference to support the DHS' policy formation in this area.
Description Evidence to select committee on social media data and real time analytics
Geographic Reach National 
Policy Influence Type Gave evidence to a government review
Impact Evidence to the parliamentary inquiry on use of data in emergencies and terms and conditions of use led to recommendations in the report regarding social media use by government and the wording of terms and conditions.
Description National policy impacts
Geographic Reach National 
Policy Influence Type Participation in advisory committee
Impact The project evidence was used in presentations to the Cabinet Office (NSCWIP) and to other government departments, such as the Home Office.