Cooperation among autonomous entities (aka agents) can take various forms. There are quite a few interaction protocols that have been standardised (e.g., FIPA: http://www.fipa.org/repository/ips.php3). Protocols are useful, but they are not sufficient to understand or model the purpose for which they may be used. A fundamental issue in cooperation is the distinction in collaborative versus competitive behaviour. Both models are used in various contexts, but it is not clear which is better for which situation and why. Simulation and validation of these models could help to understand their contexts of use.

Adaptive Decision Support

(Decision) support systems are built to support human operators in their task performance by providing them with real-time appropriate information in an easy-to-understand format. Their goal is to make operators more efficient and effective. However, most such systems are rather static and do not adapt to critical situations - exactly when they would be most necessary. Also, the filtering of information is a critical task that current systems mostly try to avoid. This results in information overload for the operators. Adaptive support systems should take the context into account with respect to information, time, individual and social situation. User and information agents need to collaborate closely and take decisions together, depending on information timeliness and operator availablity, among others. The interaction among operators can affect their performance, as well.

Dynamic Task Allocation

We like to think of hybrid communities of actors. Actors may be of organic or synthetic nature. Organic actors include human beings and animals, synthetic actors include software agents and robots. In a hybrid community the various kinds of actors need to cooperate to achieve individual or social goals. They need to allocate tasks among each other. The task allocation process is expected to be performed collaboratively and dynamically, i.e. by several actors together and depending on a current task at hand. This requires a good understanding of the properties and capabilities of different actors and means for them to communicate effectively.

Peer 2 Peer Networks

In peer-to-peer networks (p2p) the members of such a network are in direct contact without relying on any central server (except for mediation, but ideally this needs not be the case, either). The challenge is to find matching pairs or groups of network members that share a certain interest. This should happen dynamically because each member can have different interests and roles at different times. The idea to follow is that it might be a good idea to make use of recommendations of other known members. It should be possible to reach any candidate member of an interest group with six links at most. Issues in a recommendation-based mediator include trust, efficiency and completeness.

Collaborative Reasoning and Decision Making

In a multi-agent system each agent has an individual view on the world based on subjective knowledge (beliefs) that is by nature incomplete. As a society of agents tries to perform tasks collaboratively the agents need to exchange information and viewpoints. The purpose of this process is to come to a shared decision and it involves precedures for making commitments, establishing trust realtionships, making concessions concessions and reasoning about long-term benefits, among others. The goal would be to define interaction protocols and reasoning procedures that allow for collaborative reasoning and decision making under the assumptions mentioned above.

Hunters and Gatherers

The goal of this initiative was to experiment with the emergence of social complexity by starting at the beginning: primitive man. Trying to find the simple basic rules that drive primitive man would enable us to understand how more complex social behaviour could emerge. This idea is based on a paper by Jim Doran (www.davidhales.com/hugs). We still need to find some source of funding for this experiment and probably some input from an anthropologist would be enlightening.

Fishermen's Friends

I must admit that I forgot what the point was of your study of the fishermen of Katwijk. What is the question and what was the answer? Can we simulate the processes you found?

Crowd and Riot Control, Evacuation

In this TNO project we collaborate with social psychologists to create a simulation of group behaviour. The scenario deals with evacuation in a tunnel accident (car on fire). Another scenario that is planned to be implemented later on is about crowd and riot control: hooligans and their actions in relation to police presence. These simluators should lead to a platform for simulation human behaviour in situations of Fight, Fright and Flight (FFFSim).

Computer Art

The visualisation of emergence patterns of social behaviour could generate works of art of a new kind. What would this look like? Interaction with viewers would be extremely interesting to look at. Any ideas?

Cybersex

Grapje! Maybe not. When looking at the relations between humans and machines the mechanics of the reproductive act is often seen as a central theme. Look at Junggesellenmaschinen (http://www.art-service.de/article/junggesellenmaschinen_asthetik_und_naturwissenschaften_medie.html,

http://www.perlentaucher.de/buch/3308.html).

Can we use any of these ideas and apply them to crisis management, military decision making and network-centric warfare (NCW)? The trend in the defense world is to move power to the edge (http://www.dodccrp.org/publications/pdf/Alberts_Power.pdf), i.e. shift responsibility to the lowest possible instance that can deal with it. This strategy fits very well with decentralised multi-agent systems.

BDI model for reasoning. BDI stands for beliefs. desires and intentions (http://citeseer.ist.psu.edu/122564.html, http://www.cs.umbc.edu/courses/graduate/691m/spring03/papers/georgeff.pdf). Does this make any sense to an anthropologist? We are working with the University of Utrecht on the implemenation of a programming language for this model: 3APL, http://www.cs.uu.nl/3apl/. Please, have a look at the deliberation loop of the 3APL interpreter, http://www.cs.uu.nl/3apl/deliberationloop.html. This implementation is becoming part of my open source agent platform spyse (http://spyse.sf.net/).

How could we define our roles? My idea is that you come up with realistic questions and problem definitions. I could work on a simulation based on these questions. Together we would need to validate the results of the simulation (or game) with respect to the problem. Can we find customers who are interested in funding such activities? Can we write a paper together? Maybe something to submit to JASSS (http://jasss.soc.surrey.ac.uk/JASSS.html).