EVOLUTION INVOLVES SEVERAL organisms and species struggling for resources. The resulting ecosystems foster environmental changes for which adaptation is required. If real environments include other agents, and several types of interaction, from competition to co-operation, should not evaluation tasks be social by also including these agents and interactions? However, by including other agents in a test, their evaluation becomes dependent on the other agents’ abilities and their interaction in such a multi-agent system. While this looks circular, it can be turned into a recursive population-based approach, much like evolution has done. But social skills can also be seen in terms of the policy acquisition process as introduced in Part III of the book and the developmental perspective of the previous chapter. Different degrees of ‘mindmodelling’, imitation, manipulation, learning by demonstration and even self-modelling can be understood – and, eventually, measured – in this way.

WHAT IS DISTINCTIVE ABOUT SOCIAL CONTEXTS?

Social contexts are characterised by the existence of relevant interactions with other agents (conspecifics or not). Behaviours seem to be very different in these scenarios, since appropriate actions must contemplate what other agents are doing or are expected to do. More precisely, there is a continuum of contexts, being more or less social. For instance, giant pandas follow a solitary life and barely interact with their conspecifics (except for mating or marking their territory).