For me, the highlight of IROS was the Uncanny Valley special session, although the sheer size of the IROS conference and the parallel iRex industrial and service robot expo also gave much food for thought. In particular, the new coworking robots from Kawada [video] and ABB look very interesting, but it’s clear that it still takes a long time for research to transition into robust applied robotics.
The Uncanny Valley Revisited was a special tribute to Emeritus Professor Masahiro Mori, organized by co-chairs Ken Goldberg, UC Berkeley and Minoru Asada, Osaka U. Masahiro Mori’s 1970 article, Bukimi no Tani Gensho, described a phenomenon of unease that is felt as animated beings become more similar to real beings.
Labelled the ‘uncanny valley‘ in reference to Freud’s concept of the Uncanny (Unheimlich), Mori’s work first appeared in translation in 1978 and proceeded to be broadly adopted in the art world and cultural/social sciences. Astonishingly, Mori believes that robotics only noticed his concept in the mid 2000s, when papers citing the uncanny valley were presented at HRI conferences. Elizabeth Jocum from Aarlsberg U was one of several to speakers to point to the early and deep impact of Mori’s idea in other disciplines, including the arts.
It’s apt that the art world was first to appreciate the Uncanny Valley hypothesis, as Freud himself said that the uncanny is the only thing more strongly felt in art than in life. It’s also a dynamic rather than a static phenomenon, as other speakers attested. Marek Michalowski discussed the impact that animators have had on the field of human robot interaction. After all, animation has been a strong field for over 100 years and is primarily concerned with creating a compelling imitation of life. In the process, animators utilize much more than just the static superficial appearance of a character. Sound, perspective, staging, background, color and timing all enhance or destroy the illusion of life.
Freud’s original concept of the uncanny is also more closely aligned to general anthropomorphism, where the impact is less on the closeness of appearance to human and more on the human ability to ascribe emotion, agency and symbolism to logical, mechanical events or objects.
Mori intended for his theory to be a simple warning for robot designers to consider the possible affect of their constructions, and he calls for robots to be made less life-like or human-like, as he wishes for technology to have positive and not negative contributions to the world.
Robots are already out there in the world, and I think we are frequently unprepared for the range of anthropomorphism that even unlikely looking robots can attract. This is well described in the work of Nass and Reeves in The Media Equation and leads to the ‘new ontological status’ hypothesis put forward by Kahn Jr, Reichardt, Kanda and Ishiguro. Generally speaking, I find that roboticists oversimplify the uncanny valley hypothesis. Mori himself describes it as a clue rather than a theory, so it was refreshing to hear so many great speakers give it much needed dynamism and depth.