|Attentional Mechanisms for HRI - A Step Towards Building Intelligent Middleware for Socially Interactive Robots|
Seminar by Prof. João Filipe Ferreira - 1st internal seminar about CASIR project
When interacting in socially-relevant applications, robots are expected to engage with humans while displaying attentional behaviours that resemble those of their interlocutors; as a matter of fact, they are supposed to be able to assess intentionality and to be, themselves, intentional agents. Humans assess and exhibit intentionality using processes that are deeply rooted within low-level automatic multisensory attention-related mechanisms of the brain - the popular saying "there is no second opportunity to make a good first impression" is not at all accidental. It is therefore the logical conclusion stemming from this reasoning that for robots to engage with humans properly, they should also be equipped with similar mechanisms. Most of these mechanisms surface, in humans, during infancy, through interaction with their progenitors or caretakers. The acquisition of these basic skills plays a major role in the course of attaining advanced communication skills for more complex social interaction, in a process that takes place throughout the first twenty-four months of a child's life. This fact has impressed the idea on many robot designers and researchers in artificial intelligence (AI) that robots should follow a similar route, engaging in sessions of what is called learning by demonstration to acquire equivalent mechanisms, from a "blank slate" starting point. This is part of the epigenetic and developmental perspectives to robot design and AI. However, we argue that, although the origin of the robot's basic social skills is certainly relevant, even more important is the potential of an emergent property of more complex skills from basic building blocks, irrespectively of if these skills are learnt or preprogrammed. In our opinion, socially interactive robots would greatly benefit from the development of probabilistic real-time frameworks implementing automatic attentional mechanisms. These would effectively constitute middleware for the development of more intelligent and complete socially interactive robotic systems. In fact, we argue that this hierarchical/modular architecture-oriented approach would equip socially-skilled robots with fundamental innate and preacquired capabilities that humans already possess from a very early age; it is our belief that failing to do so will most probably ultimately result in robots exhibiting social deficits similar to those of individuals suffering from autism spectrum disorders. With this in mind, this presentation will attempt to paint a picture of the current situation concerning social and service robotics, how artificial cognitive systems with attentional capabilities fit in this picture, and what challenges might lie ahead for roboticists who might want to pursue this line of research.
J. F. Ferreira and J. Dias - "Attentional Mechanisms for Socially Interactive Robots - A Survey". IEEE Transactions on Autonomous Mental Development, Volume 6, Issue 2, p. 110-125, 2014. DOI: 10.1109/TAMD.2014.2303072 .