Terrorism and Game Theory*

Coalitions, negotiations and audience costs


The aim of the research was to investigate theories of terrorism and its changes over time and apply this to game theory, in particular the concept of audience costs and coalition formation within the game theoretical framework.
Causes and goals of terrorism were identified and modifications of ideas of terrorism in time observed, accompanied with an analysis of involved actors.
Based on a premise of justification derived from the Just War theory, negotiations with aggrieved groups (terrorists) can be a rational approach to such type of protracted conflicts and uncovered dynamics based on the composition and character of organised actors. This includes averting the joint bargaining paradox when asymmetric actors cooperate to achieve a joint goal, governments who may fare better from their point of view when dividing up refugees into geographically separate areas to decrease the possibility of cooperation between aggrieved groups and recognising two-speed negotiations, which can have a (detrimental) effect on (deadline) negotiations. Secondly, an adaptation of the audience cost model explained the "war of nerves" of the terrorist theatre involving the generation of fear amongst its targets and the updating of information on moderate actors by action of extremists. Several mutations of the Prisoner's Dilemma were discussed as options to prevent mutual harm.


The conclusions are categorised in accordance with the research questions as outlined in chapter 1. The investigation uncovered the following answers:

What theories of terrorism do exist and are they still of relevance today and/or have they changed over the past 25 to reflect changes in the global society?
Results of scholarly research over the decades are inconclusive about the concept of terrorism and its causes and goals; there exist a plethora of theories of terrorism, partly due to changes of the subjective idea of the (academically) unclear meaning of terrorism, the modifications of actions, but not tactics, carried out and because of difficulties investigating aggrieved groups. Theories have changed over time, and likely will change in the future, reflecting changes in the global society.
Causes range widely from ethnicity and nationalism/separatism to poverty, economic disadvantage and globalisation. In addition, (non)democracy, Western society, disaffected intelligentsia, dehumanisation and religion may be connected to instigating terrorist activities as well, but none of the aforementioned 10 potential causes has a single conclusive causal relation: it is possible to devise arguments both confirming and refuting the connections. Likewise, goals vary, and include demands for varying levels of power and/or territory, implementing a certain ideology and advocating religion. As a result, the emergence and composition of aggrieved groups (terrorist organisations) and the interplay with other actors (states, groups and international organisations) do not follow one specific model either.

If the answer is yes to the previous research question, can be devised why theory / theories did change and can be identified what and when changes occurred?
Whereas in the 1960s and 1970s ideology-based terrorism was more prevalent, it was narrowed down to the Red Network as main instigator by the mid 1980s with a counter-reflex of research into state-sponsored terrorism, i.e. the West as main perpetrator, in the late 1980s and early 1990s. From the late 1990s to present, religion as cause for resorting to terrorism, and to some extend the First-Third World dichotomy, receives disproportionate attention and dehumanisation of perpetrators does take place in the popular literature. In addition, a shift towards Freizeit-terrorism and cell-based organisation structures on a supra-national level can be identified. Near-future devastating 'superterrorism' has been predicted for, roughly, the last 15 years, but has not materialised even though the technology to do so is available. However, new tools do not imply a more violent mindset of members of an aggrieved group.
Theories, and varying emphasises on their causes, tend to shift as a result from changes in the international political arena, and to some extend receive disproportionate attention when domestic conditions of the countries fighting 'terrorism' are worsening to divert attention.

Determine which aspects of Game Theory may be useful as an aid in modelling activities surrounding dealing with actors involved in terrorism.
Variable-sum normal form games, like mutations of the Prisoner's Dilemma can capture basic elements of negotiations between aggrieved groups and the government and an adjusted War of Attrition is suitable for modelling audience costs. Extensive form games disclose interactions between actors in a structured format. Cooperative games reveal interesting features of coalition-formation and strengths and weaknesses in negotiation processes.

Using the model(s) of a theory of terrorism, test the validity of, and adjust where appropriate, different game theoretical models of the involved actors in terrorism.
· Joint bargaining paradox and coalition forming: is there an optimum in the amount of factions and type of coalition-forming? Can they be expected to be stable and fruitful, and if so, when?
In a 3-player game, where two players consider forming a coalition, this can be favourable in situations where the fallback position is lower than in a non-cooperative game, provided that the two coalition players divide the bargaining gains asymmetrically and both agree on this asymmetric subdivision ratio. I have proved that a strong moderate terrorist can fare well by cooperating with a weak (smaller) extremist faction (the same holds for the case of one terrorist (representative) and two government players of unequal strength). Further, if the sets of means and goals of the two aggrieved groups are sufficiently close substitutes, the equilibrium form of organisation is an encompassing group; if they are sufficiently complementary, the equilibrium form of organisation is in separate groups. When the former is applicable, a government would benefit from the 'divide and rule' tactics by distributing refugees into smaller groups in geographically distinct areas to avoid cooperation against an oppressor.
Deadlines have a negative outcome on peace negotiations and processes; the discussed models provide a game theoretical explanation why this is observed in the field based on the internal dynamics of the actors (organisations/government). This also indicates that when one can identify a situation with 'two-speed' actors, based on the organisational or institutional arrangements involved, it may alleviate some of the frictions and aid mutual understanding for their respective inner workings, hence aid towards a positive outcome of negotiations.
· Government negotiations with terrorists.
This is partly addressed and explained in the previous point. The experimental game did not provide sufficient statistical evidence to support the claim of a division between moderates and violent extremists, but was observed during one game out of three.
· Audience costs: to what extend would it be possible to adapt the concept of audience costs to a terrorism framework, with regard to the 'terrorist theatre' and democratic states desiring to 'combat terrorism'?
In terrorist frameworks like peace negotiations, audience costs can be generated and identified, in the non-negotiation phase aggrieved groups exploit the audience cost model to their own benefit, alike a War of Nerves. Modelling audience cost parameters, especially the rate of deduction in crisis prolongation, depends on the problem being modelled and the (subjective) preference, or moral bias, of the modeller.

Does the application of Game Theory on terrorism model(s) provide new insights, which might aid towards not only an understanding, but also provide ideas towards a possible resolution of such type of conflicts?
Game theory is a useful tool in rationalising the emotion-laden field of terrorism, and has provided insight in the intricacies of the audience cost model, the increase of violence during peace negotiations, options to escape the mutual harm of the Prisoner's Dilemma, potential for coalition-forming and offers an explanation for unconstructive deadlines and actors in two-speed negotiation processes. Although several of the uncovered aspects will need to be verified with empirical data, at the present stage it already aids understanding, which is a first step towards resolution of conflicts, but one can never include all terrorism-related aspects into one model because of the controversies surrounding the concept 'terrorism'.

Suggestions for further research

Although the conducted research did clarify aspects of the terrorist theatre, it also opened new areas one can explore for further research.
English-language based sources are relatively one-sided on terrorism and do not capture all angles. Even within the Western views, probed continental European material emphasised other aspects of terrorism not, or hardly, touched upon by the English scholarly literature, therefore a comparative investigation in differences of point of view and reporting between countries and across languages could reveal additional perspectives.

I addressed the game theory of cooperative structures with regard to internal group dynamics related to unanimity and majority positions on a theoretical level, which would benefit from a closer analysis of terrorist organisations and cells to put the ideas to the test on how they reach a policy stance and if it indeed affects negotiations in the way as predicted by the theory. Possibly related are aspects involving the logic of collective action, peer pressure and internal motivation of the terrorist. Overall, this could shed light on deadline- and two-speed negotiations and sustainability of a peace agreement by their grass roots, whether from the aggrieved group, government or the wider public.

The audience cost model could not be used to assess build-up of audience costs within aggrieved groups due to a lack of sufficient information, though it would be highly informative if Crenshaw's (1991) idea of organisational disintegration were related to a leader who incurred audience cost due to making false promises to his group members could be tested and how these audience cost are 'paid' when there are no elections, or if there are other more important reasons.

The dynamics of inter-group and state relations is underexposed. Although several examples of negotiations exist, at present most notably Northern Ireland and Israel/Palestine, there is a strong US voice to be 'tough on terrorism' (despite existing 'quiet diplomacy' of negotiations with aggrieved groups, e.g. USA with Hizbollah), which does not aid in 'openly' researching the matter. However, a better understanding of these dynamics could help actors involved in other protracted conflicts to set out their respective policies. These envisaged case studies could provide information in order to determine probabilities of the extensive form games to update either the mixed strategy or typing of players via Bayesian updating, which in turn serves (more accurate) prediction of behaviour in these conflict situations. Ouardighi's (2003) model of trust should be put to the test, i.e. one can asses the various conflicts on this dimension, which could reveal if requests by one faction to "just trust us" and another choosing vigilance is rational or purely emotion-based. Additionally, it might reveal if external monitoring fosters trust, or if it is indeed harmful as Ouardighi devised.

The idea of a Coalition Calculator is promising proof of concept based on results obtained from the experimental game, and deserves further attention in looking into its possibilities to predict potential for coalition-forming / defection. This includes further development of the software and, ideally, being put to the test in real negotiations, which can be various low-intensity conflicts as well as other settings like GATT negotiations.

Several improvements on the experimental set-up can be made. These include informing the players on "terrorist 1 and terrorist 2" as opposed to "a moderate and an extremist", using several groups in parallel with each an observer and/or the game is tape recorded in order to capture the bargaining dynamics. Other improvements may be a more restrictive setting with alternate bargaining and introduction of a 'punishment' factor (deduction in payoff) for instigating violent events, although the latter is an entirely moral consideration. Last, a note of caution on modelling a conflict, is that due to the sensitive area, one may not expect that all participants can detach themselves from the real-life conflict the game is modelled on and this prejudice can affect their negotiation behaviour. Therefore, it may be more effective to model less well-known problems.

Download the full report (in PDF format, 763kb) and view related publications.

* This 7 months MA dissertation (grade: A) was carried out in 2003 at the Department of Government and Society, University of Limerick, Ireland.