I would like to conclude by briefly proposing a perspective for future research based on the system reaCog. As presented, its ability to interact and cooperate with other agents is fairly restricted. At the same time, the pre-requisites of a broader social extension of the system seem to be in place. The present paper already shows how reaCog could be equipped with the capacities to recognize the behaviour of others and apply a Theory of Mind. In their 2011 paper, Cruse & Schilling further propose that by implementing a two-body model (a “We-model”) reaCog might be capable of cooperative behaviour using shared goals. Integration and further expansion of such social capacities, and their application in an actual robot, seems promising considering the importance of social interaction in processes such as language acquisition and emotional regulation. Some have even suggested that the presence of other agents in the environment, or, in other words, dealing with social complexity, was a dominant factor in the evolution of sophisticated cognitive abilities (Humphrey 1976). Thus bio-robotic research in this direction might provide new insights into the mechanisms underlying such developmental and evolutionary processes. Moreover, a social extension of reaCog might eventually shed light on potential emergent phenomena on a group level, such as labour division, collective planning, social hierarchies and, most fundamentally, joint action coordination. What high-level social phenomena emerge when multiple bio-robotic systems like reaCog interact with each other?
Cruse and Schilling’s system seems particularly well-suited to further illuminate motor theories of social cognition. According to such theories, the important social cognitive capacity of understanding another’s actions is directly linked to mechanisms that are active when the observer performs similar actions Gallese et al. 2004; for criticism see Jacob & Jeannerod 2005). The underlying neural mechanism has come to be known as the mirror-neuron system. Furthermore, there is evidence that the mirror-neuron system plays a role in certain aspects of self-consciousness. For instance, Uddin (2007; see also Molnar-Szakacs & Uddin 2013) suggests that this is the case for representations of the physical self, and ascribes frontoparietal mirror-neuron areas an important function for self-recognition (especially the recognition of one’s own face). As mirroring mechanisms can be integrated in reaCog as well, this opens the possibility of further investigating motor theories of social cognition and the relation between internal motor simulation and the self in a quantitatively defined system.
An ability that is highly important for human social interaction is the ability to communicate using language. At this point, the linguistic capacities of reaCog still seem quite inflexible and limited in scope. A highly interesting extension of this system would be to provide it with the means to learn words and their meanings by interaction with other agents. Some of the pre-requisites, like the ability to internally simulate the behaviour of others, could, as Cruse and Schilling argue, be implemented in reaCog by using its internal body-model to represent another agent. Robotic research in this direction was performed by Steels & Spranger (2009). Their artificial systems are capable of autonomously acquiring a simple language consisting of words for specific body postures. After learning is complete, the artificial agents are able to reliably assume body postures on verbal command by other agents. Since social learning has also been implicated in the process of concept formation (Steels 2002), the proposed extension might also foster our understanding of this intriguing phenomenon.