Awareness as a vital ingredient of teamwork

For successful coordination and cooperation in a multiagent system, participants need to establish a sufficiently accurate awareness of the current situation. Awareness is understood here as a limited form of consciousness: in the minimal form, it refers to the state of an agent's beliefs about itself, about others and about the environment. When considered in the context of agents' mental states, this leads to distinguishing three levels of awareness: intra-personal (about the agent itself), inter-personal (about other agents as individuals), and group awareness.Problems in modeling agents' awareness on all three levels are analyzed. It turns out that both the communication medium and agents' cognitive and computational limitations make the achievement of awareness difficult. Cognitive science is used to analyze and explain problems in human awareness, based on the concept of bounded rationality. The BDI framework, originally designed to formally define agents' informational and motivational attitudes, turns out to be also fruitful both for precisely formulating the problems concerning agents' awareness, and, even more importantly, for formulating and comparing possible solutions. Thus, the two fields of cognitive science and MAS mutually benefit from each other's viewpoints, especially in the light of the currently growing need for teamwork in which both computational agents and humans are involved.In this paper, some possible avenues to solutions for defining and achieving appropriate levels of awareness are suggested. In some cases, these are concrete formal solutions, which have been adopted in our theory of collective motivational attitudes, presented in a number of conference and journal papers [5, 6, 7]. They give rise to more generic solutions that can be of use in any advanced BDI system, especially in those aiming to realize teamwork.