What Do We Learn When We Learn by Doing? Toward a Model of Dorsal Vision

摘要

Much effort in computer science is currently focused on developing architectures for multi-agent adaptive system capable of monitoring the environment and detecting security threats. I present here one such architecture developed by evolution and implemented in the neural mechanisms of the human brain - it is the dorsal visual system. I claim that the dorsal visual system in the human brain can be modeled as two cooperating rough agents which monitor the environment and guide other systems. The two agents' adaptation capabilities can be modeled on the basis of research in neuroscience related to the processes of implicit learning from experience. In the paper I first present arguments behind my claim. Next, I show how studying the dorsal visual system may help to improve human-machine interaction. Finally, I suggest how the conjectures presented here can be tested experimentally.