Keepon the Robot Interacts without the Overload

Hideki Kozima and Marek Michalowski (programmer) are creators of the robot Keepon, which has danced to Spoon in an outrageously popular YouTube (below), reprised in a professionally produced film on Wired's dollar, and, yes, has worked for years in autism research and therapy. Keepon is a simplified robot specialized in nonverbal social behavior, developed for interaction with (all) children.

Keepon takes in some ways the opposite approach of Transporters (also, here)--it reduces and simplifies, removing the alarming complexity of a human face and distilling nonverbal social behaviors down to their most basic level. Keepon's design deliberately avoids the uncanny valley problem that made me find Transporters so frightening, which Michalowski discusses specifically in the Post-Gazette article with his comparison between the greater popularity of cartoony CG over photorealistic CG in movies.

More fascinating--and exciting--to me though was looking over some of the actual papers published by Kozima on Keepon's interactions with autistic toddlers as the toddlers began engaging in spontaneous social interactions with the robot. From Social interaction facilitated by a minimally-designed robot: Findings from longitudinal therapeutic practices for autistic children,

Our major claims are (1) simple robots with minimal and comprehensive expressiveness facilitate the spontaneous exchange of mental states in autistic children, (2) autistic children therefore possess the motivation for this mental exchange, and (3) the major social difficulties that autistic children generally suffer from would rather stem from the difficulty in sifting out socially meaningful information (e.g., attention and emotions) from vast incoming perceptual information.

In other words, the results of the Keepon study suggest that autistic people aren't lacking the ability to comprehend other minds or without social capacity or motivation as some theories suggest, but are having difficulty making sense of the intensely complex wash of social information found in more human social partners. This idea is actually completely consistent with my own experience of trying to manage social interactions--in my language, they are always "too bright."