My agents might be very different to your agents as I work in Computer Science and focus more on Social Simulation. I define archetypes and provide every agent with its own personality based on these archetypes. This influence the likelihood to reach certain states or the time the agents spend in certain states.

You can find an example here:

http://jasss.soc.surrey.ac.uk/14/2/2.html

Our latest research project is about simulating peace building processes. Here we consider psychological factors like anger and fear. I could imagine you could also have a factor like ethics that might be influenced by anger and fear.

http://www.cs.nott.ac.uk/~pos/projects/simpb/

In the bottom of the following page you find the agent definition.

http://www.cs.nott.ac.uk/~pos/projects/simpb/screenshots.html

You might find some more interesting stuff on my website, e.g. our attempts to bring Economists and Computer Scientists together to discuss the application of Software Engineering tools to define Economics ABMs.

http://www.cs.nott.ac.uk/~pos/bemoossig/

Happy to discuss more if you are interested.

I suggest you go back to the work of Axelrod in his book, _The Evolution of Cooperation_. In there you'll find how different strategies compete and compare with one another. Additionally, he describes different "ethical" strategies and how they perform.SPOILER: the cooperative agents perform the best.

http://books.google.de/books/about/The_Evolution_of_Cooperation.html

Take  a look at:

Castelfranchi, C., Dignum, F., Jonker, C. M., & Treur, J. (2000). Deliberative normative agents: Principles and architecture. In Intelligent Agents VI. Agent Theories, Architectures, and Languages (pp. 364-378). Springer Berlin Heidelberg.

A basic principle in this type of problem is the successful implementation of Deontic Logic

You may wish to include Assumption Based Truth Maintenance as a constraint.  State your 'ethical assumptions' and impose on all agents.

Perhaps, most abstractly, unethical behavior involves doing things that bring about undesirable states of affairs and not doing certain things that bring about desirable states of affairs.

 thanks you all for your contributions.

I found very interesting the tracks and the ideas you have proposed.

My agent is actually different from yours. it's consists of a constraint-based local solver and a collaborative communication protocol, allowing to have a comprehensive and consistante solution compared to inter and intra agents constraints. However, an agent may have an unethic behavour (lie, ....) and provide a local solution, preventing the system to converge to a globale effective solution or mislead the system to make wrong decisions.

As well as ATM (previously mentioned) you may wish to look at Samarandache's work at

http://www.gallup.unm.edu/~smarandache/neutrosophy.htm

in order to negate/manage agent's unethical behavior

Yes, I found your first track very interesting, I'm now focusing on the work Johan de Kleer. An assumption-based TMS. Artificial

Intelligence, 28:127–162, 1986 and Extending the ATMS. Artificial Intelligence,

28:163–196, 1986.

Regards

Wouldn't that also depend on the goals of the cooperative environment?

There is no clear answer to your question, since defining "unethical behaviour" is not straightforward (what might be unethical for some, might be ethical for others).

In any case, one way of dealing with unethical or bad-behaving agents is to allow the agents (individually and/or collectively) to somehow monitor and evaluate the behaviour of the other agents (e.g. self-monitoring, self-evaluation). This would allow the system to detect such bad-behaving agents (e.g. free-riders, malicious/lying agents) and react accordingly, for instance by not considering their solutions (in your case of constraints solving), excluding them from the system, imposing some kind of penalty, etc. 

De Kleer is a good start - some additional pointers:

Coordination for ATM

You may also wish to develop a variation on Chang et al.’s. (1993) research on distributed reasoning supporting multi-auditor cooperation. The auditors can be represented as agents. The Chang et al. presentation addresses the support for a team of cooperating auditors.

During the cooperation stage, the auditors must ensure that their default assumptions are not contradicted by the empirical evidence of other team members. The cooperation process requires a narrowing down of the areas of conflict, identifying the areas where further testing is required and the development of the explanation for the consensus opinion that emerges from the process (Chang et al.1993, p. 347).

As part of their analysis, auditors develop propositions or beliefs, based on their assumptions . These propositions may then be communicated to a more experienced auditor who may judge them as true, false or unknown (this triplet may be seen as a component of the linguistic). The judgments of the more experienced auditor are subsequently communicated back to the originating auditor who negotiates until a consensus is reached.

Chang, A., A. Bailey Jr. and A. Whinston (1993). "Multi-Auditor Cooperation: A Model of Distributed Reasoning". IEEE Transactions on Engineering Management, 20(4): 346-59.

[1]These assumptions are by nature default assumptions, which hold that in

   the absence of evidence to the contrary, the item under review is sound.

@Peter, thank you for your additional explanations, I find it valuable for my investigations. Are there any more new interesting paper in this topic ?

@Didac, I competently agree with you on this way of dealing with unethical or bad-behaving agents. Thanks for contribution. 

defining the ethical and unethical is very difficult may be for those agent things are ethical but for others it is not, just like the human nature in the society some people support an idea and say it is best but some people oppose it. so a lots of confusion  for this ethical and unethical terms. but for any particular case we set the limit or constraints to the agent for learning. because we develop the agent for a particular environment so we use a separate agent for that or set the degree of learning.

Ethical vs Unethical behavior imho is a very difficult aspect to meadure because ethics depend on the domain, the culture and observer of the agent (s) interactive.  If we assume a Cooperative Scenario with Benevolent Agents and defining an ethics code according to a set if social rules accepted by observers then plausibly it could be done.  Not sure... It is a very soft matter due to the fact that even with humans ethical behavior is relative to what ethics code is follwed.  In other hands, great research topic.