The property of agent's sacrifice: definition, measure, effect and applications
Viacheslav Abrosimov · International Journal of Reasoning-based Intelligent Systems · 2016
This paper studies situation when an agent fulfilling a collective task in a multi-agent system performs actions inconsistent with its further functioning. Such a state reflects the agent's capability for self-sacrifice. The belief-desire-intention architecture of multi-agent systems is extended by adding the notion of the agent's 'commitment' to a group and the emotion of the agent's 'sacrifice'. A measure of sacrifice is defined as the sum of fuzzy numbers characterising the agent's commitment to the group, the degree of the agent's belief in the reasonability of self-sacrifice and its intention to resolve the collective task. Certain conditions under which an agent becomes a sacrifice for collective task fulfilment and a logical sacrifice selection procedure based on negotiations among agents are established. Finally, the behavioural models of agents with different self-sacrifice measures ('altruists', 'pragmatists' and 'egoists') are constructed using AnyLogic simulation tools.