Non-invasive Estimation of Stress in Conflict Resolution Environments Paulo Novais, Davide Carneiro, Marco Gomes, José Neves DI-CCTC University of Minho Largo do Paço, 4704-553 Braga, Portugal {pjon; dcarneiro}@di.uminho.pt, pg18373@alunos.uminho.pt, jneves@di.uminho.pt Abstract. The current trend in Online Dispute Resolution focuses mostly on the development of technological tools that allow parties to solve conflicts through telecommunication means. However, this tendency leaves aside key issues, namely our concern with respect to context information that was previously available in traditional Alternative Dispute Resolution processes. The main weakness of this approach is that conflict resolution may become a cold process, focused solely on objective questions. In order to overcome this inconvenience, we move forward to incorporate context information in an Online Dispute Resolution platform. In particular, we consider the estimation of the level of stress of the users by analyzing their interaction patterns. As a result, the conflict resolution platform or the mediator may weight to what extent a party is affected by a particular matter, allowing one to adapt the conflict resolution strategy to a specific problem in real time. Keywords: Multi-agent systems, Online Dispute Resolution, Stress, Influence Diagram Model. 1 Introduction Online Dispute Resolution is now seen as the new technology-based paradigm for solving disagreements, and then replacing litigation in court. However, as the human’s role gradually loses its substance as the main decision maker, some elements must be taken into consideration, so that conflict resolution processes guided by autonomous software agents will incorporate the best facets of the human experts [1]. Concerning interpersonal communication, Mehrabian [2] points out that the non-verbal elements are particularly important for communicating feelings and attitudes. In that sense, the use of technological tools for communication, with the consequent separation of the interlocutors may represent a risk, as a significant amount of context information is lost. It is our conviction that these issues should be taken into consideration when developing technology-based conflict resolution platforms. Specifically, we believe that the most suited approach merges insights from Multi-Agent Systems (MAS) [11] and Ambient Intelligence (AmI) [3, 10], which we have applied previously with success in other domains [12, 13]. Our objective is to use intelligent environments to support the conflict resolution process. Basically, we are extending the traditional technology-based conflict resolution model, in which a user simply interacts with the system through