The emergence of specialization remains a true challenge. Suppose a world initially filled of specialized and unspecialized agents and let's define these later as able to play the various parts endorsed by individual specialized agents (and to play these parts just as well as the specialized agents, un-specialization would be here synonymous with "multi-specialization"). In an evolutionary perspective, it is trivial to admit that a specialized agent will always appear as less adapted than an unspecialized one, which can indeed alternatively plays the many roles. The unspecialized agents will meet much more agents (specialized and unspecialized) and much more situations to which they are adapted too and so cumulate much more payoffs (unless we "penalize" the payoff of un-specialization). This indeed appears as a paradox to face in order to make sense of a world nonetheless full of specialization. Since such an issue could be of great concern to the researchers of IRIDIA interested in collective behaviours executed by distinct agents, I propose in this talk to raise the challenge (if any.We could conclude that this is a "non-problem") and animate the discussion, rather than delivering a clear and clear-cut solution to that problem.
Specialization, Multi-agent systems