Abstract

What is an agent? What are the essential features that differentiate an agent from other biological and artificial systems? In this paper, we flesh out a series of theoretical obstacles that hamper the naturalization of agentive self-awareness - for one, the use of covert mentalistic notions like action. To overcome these issues, a set of possible solutions is taken into consideration: a cognitive intermediate level, development and situatedness, externalism, intrinsic motivations, and bio-inspired talamo-cortical models. Eventually, a robotic architecture based on such approaches is briefly described and discussed.