© Copyright JASSS

  JASSS logo ----

S. Thoyer, S. Morardet, P. Rio (2001)

A short answer to Bruce Edmonds's commentary on

Sophie Thoyer, Sylvie Morardet, Patrick Rio, Leo Simon, Rachel Goodhue and Gordon Rausser (2001)

A Bargaining model to simulate negotiations between water users

Journal of Artificial Societies and Social Simulation vol. 4, no. 2,

To cite articles published in the Journal of Artificial Societies and Social Simulation, please reference the above information and include paragraph numbers if necessary

Received: 27-Mar-01      Published: 31-Mar-01

* Introduction

According to Bruce Edmonds's criticisms, our paper is supposedly plagued by three main faults :
  1. the model is based on unjustified assumptions concerning the rationality of players (perfect information and substantive rationality) and the real negotiation process (the disagreement outcome, the backward induction structure) .
  2. the model does not provide any new insights into the real negotiation process since the outcomes of the simulations reflect the structure of the model and do not reveal any new counter-intuitive behaviour which would not have been explicitly designed in the model.
  3. The model has no predictive power and it cannot be helpful as a negotiation-support tool since it is based on false assumptions and since it does not allow a better understanding of the negotiation issues at stake.
Beyond technicalities, these three points sum up in one single argument: the inadequacies between our game theory model and the reality. In short, Bruce Edmond regrets that our efforts to research into relevant facts concerning the case study be wasted by the use of a fallacious "traditional technique".

No easy answer can be offered to Bruce Edmond because his comments raise a tricky question which is treated differently in the game theory field and in the multi-agent modelling field: do we need a theoretical basis to formalise an empirical problem and translate it into a workable model? Or do we just need to reproduce as adequately as possible the nature of the problem?. It leads us to develop three points: (i) the construction of formal approach?; (ii) what would be a more realistic model?; (iii) competition or complementarities?

* The construction of a formal approach?

A theory can be interpreted as a cumulated sum of knowledge and know-how built on a solid internal coherence. It provides an analytical framework to tackle empirical questions. Game theory is based on coherent concepts concerning the nature of negotiation solutions and the properties of equilibrium solution and it allows us to link them to other economic concepts such as the Pareto-efficiency of a solution.

Therefore the construction of the formal approach takes place simultaneously in the theoretical domain and on the field. It must mimic the main characteristics of the issues at stake within the set of coherence constraints imposed by the theory. However, it implies that a reasonable trade-off be accepted between a fine description of the reality studied and the coherence of the underlying concepts (an excessive ambition to reproduce the reality may mute the model by locking it definitely within the requirements of its local validity).

* What would be a more realistic negotiation model?

We argue that our multilateral multi-issue bargaining model has the capacity to capture a large scope of the salient features of a real-life negotiation process: agents can hide their objectives and manipulate others, they can form coalitions, they can ask for an enlargement or a reduction of the negotiation space, they are ready to accept compromises on certain negotiated variables in order to obtain more on others.

Our main weakness was rightly pointed out by B. Edmonds and by other anonymous referees and concerns the lack of credibility the perfect information assumption. We acknowledge that the model would be greatly enriched by making endogenous the information structure (in particular the communication and the learning procedures between agents) as well as the negotiation space (which is built up progressively through the development of a common shared vision of the issues at stake).

* Competition or complementarities?

The objectives of our game theory model are quite similar to the objectives of most multi-agent systems: improve the understanding about the characteristics of the negotiated solution and develop a negotiation-support device. It would be disappointing to miss an opportunity to enrich our respective tools due to an initial misunderstanding. We have already sized the opportunity by launching a new research programme in France which gathers sociologists, game theorists and multi-agent system specialists. Our explicit and common objective is to build an integrated methodology which would allow each analytical approach to benefit from the research outcomes of others.

ButtonReturn to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, 2001