Seminar: Understanding Minds through Synthesis: A Neuro-Robotics Study
The next L&M seminar in Japan will take place on Monday, Nov. 14, at 4pm, in Room 107, Building 16, at the Komaba campus of the University of Tokyo. We are pleased to have Jun Tani from Riken BSI in Tokyo.
Understanding Minds through Synthesis: A Neuro-Robotics Study
Behavior and Dynamic Cognition Lab
Riken Brain Science Institute
The difficulty in understanding minds is that their problems are inherently multifaceted. We can delve into so many different aspects and issues of minds, including subjectivity and consciousness in phenomenology, behavioral correlates in psychology, computational metaphors in cognitive science, biological implementation in neuroscience, and pathology and mechanisms of mental disorders in psychiatry. A good model of minds should, therefore, account for all these different aspects and would be better than having a different model for each.
In this talk, I introduce the challenges in modeling minds that I and my colleagues have been attempting to solve as part of a neuro-robotics research project over years. We have proposed, as the core model in the project, a “generative model” with hierarchy which features two essential processes: a top-down prediction process and a bottom-up regression process for perceptual sequences associated with actions. In the top-down process, perceptual flow of visuo-proprioceptive sequences can be predicted and projected into the lower peripheral level as biased by intentional states appeared in the higher level. On the other hand in the bottom-up regression process, the error signal generated between the prediction and its perceptual outcome in the lower level propagates to the higher cognitive level and this signal modulates the current intentional state in the direction of minimizing the error which results in adaptions of the system to possible situational/environmental changes.
In our neuro-robotic experiments, implemented with the aforementioned model, we could observe various nontrivial phenomena especially in dynamic interactions between the top-down prediction and the bottom-up regression processes. By making possible correspondences of these observed phenomena to cognitive, phenomenological and psychiatric aspects of minds, I discuss what we can gain in understanding minds.