Tomorrow (11 June, 2013) I am giving a seminar at the Bristol Robotics Laboratory (BRL). It is five years since I last visited the lab, and I’m looking forward to seeing how the projects I saw back then have developed, as well as getting the opportunity to see new projects that have started more recently.
This is the title and overview of what I have planned for the seminar:
“Tempered” Anthropomorphism and/or Zoomorphism in Human-Robot Interactions
In this talk I consider how a level of “tempered” anthropomorphism and/or zoomorphism can facilitate perceptions of, and interactions between, overtly different communicators such as humans and non-humanoid robots.
My argument interrogates the tendency within social robotics simply to accept the ascription of human characteristics to machines as important in the facilitation of meaningful human-robot interactions. Many
scientists and other academics might argue that this decision is flawed in a similar way to scholarship that attributes human characteristics to animals. In contrast, my analysis suggests that it is possible to adopt
a “tempered” approach, in particular when the robot other is overtly non-humanoid. I suggest that a level of projection is unavoidable, and is quite possibly the only way to attempt to understand autonomous or
semi-autonomous robots. However, being constantly reminded of the “otherness” of the machine is also vital, and is of practical value in creating effective multi-skilled teams consisting of humans and robots.
I will try to alter my talk as required in response to my audience’s reactions, since it can be quite challenging to present humanities-type research to a predominantly technical audience. My aim is to emphasise the practical use of reconsidering human-robot interactions in this way.
Baxter is a new workplace robot developed by a company called Rethink Robotics, run by Rodney Brooks (a very well known robot designer, who also started the company iRobot famous for developing Roomba robotic vacuum cleaners). Baxter is designed to work next to people in manufacturing environments, being human-like in form and consisting of a torso and two arms, together with a screen “face” (well, one consisting of eyes and eyebrows at least).
Most interesting to me is the way in which people communicate with Baxter, using touch first to get the robot’s attention and then to move the robot’s arms into particular positions. This reminds me of the touch of a yoga teacher, for example, in helping to position people into a particular pose. Baxter also has a graphical interface, displayed on the same screen that more often shows the robots eyes, which is controlled with buttons on each arm. In order to “program” Baxter to complete a task a person can therefore show the robot what to do by moving its arms into position and choosing the appropriate action, eg to pick up or place down a part, from the user interface. Importantly, it is the way in which Baxter learns a movement from being placed into position that seems to separate it from various other manufacturing robots currently in production.
As Rodney Brooks explains in the interview with IEEE Spectrum writers Erico Guizzo and Evan Ackerman, “[w]hen you hold the cuff, the robot goes into gravity-compensation, zero-force mode” such that “the arm is essentially floating”. This makes the robot easy to position, and as Mike Bugda notes in the video below, Baxter is therefore understood to be “very compliant to the user”. Although “compliant” is used here in part to emphasise that the robot is flexible and therefore able to deal with “an unstructured environment” (Matthew Williamson, in conversation with Guizzo and Ackerman), there is also a sense in which this robot is being placed as a servant, or possibly even a slave, by virtue of its immediate compliance to a human’s touch. This design decision in itself is probably a pragmatic response to making the robot easy to program in the workplace, but from my perspective it raises some issues since this is also clearly a robot designed to be read as human-like, but also as a compliant servant/slave.
The idea of Baxter as human-like is reinforced when Bugda (in the video below) explains that Baxter “exhibits a certain level of interaction, collaborative friendliness that makes it easy to understand and introduce into a manufacturing environment”. For example, when you touch its arm the robot stops what it is doing and looks towards you, acknowledging your presence, and once its instruction is complete the robot acknowledges that it understands with a nod of its “head”. Guizzo and Ackerman take this idea further, when they suggest that while at work Baxter’s “animated face” displays “an expression of quiet concentration”, while a touch on the arm causes Baxter to “stop whatever it’s doing and look at you with the calm, confident eyes”.
Although this video is simply a demonstration, Baxter has obviously had some limited field testing, and Brooks (again in conversations with Guizzo and Ackerman) notes that after people have worked with the robot for a while “something interesting happens … People sort of personify the robot. They say, ‘It’s my buddy!’”. It is at this point that the perception of the robot as a friend is reinforced.
This type of reading of Baxter as a “buddy” or friend might be assumed to be closely linked to the robot’s human-like form. However, my research considering the ALAVs, the Fish-Bird project and also Guy Hoffman’s robotic desk lamp, AUR, along with anecdotal evidence from friends who are owners of Roomba robotic vacuum cleaners, indicates that robots of pretty much any form encourage this kind of personification. In addition, for Baxter, I suspect that the use of touch between human and robot might also serve to support the perception of this robot as a social subject, and eventually a friend. The importance of touch in working with Baxter would seem to sets this robot apart from others that I have considered in my research to date. This might also suggest that Baxter could be made more machine-like (and less human-like) in a move that would reduce my discomfort in placing humanoid robots as servants/slaves, as I have suggested occurs with this robot worker.
Although I have written about things other than robots before here, it has been a long time since I wrote about anything. When I finished my PhD I resolved to blog again, but, of course, that isn’t what happened. Instead, I had some time off and then threw myself into sessional teaching. There have been benefits, for example my decision to do that allowed me to apply for and gain an Early Career Development Fellowship at Curtin, but I since the beginning of this year I have found, or made, no time to blog.
Today I’m trying something new. Although I will often still try to find images to pepper my posts, and I will continue to write about robots from time to time, I am freeing myself up to improvise a bit more in this space. In part, this is because yesterday evening I reminded myself that it is possible to stand in front of a small group of people and improvise a talk. If I can do that, then I think it might also be a good idea to learn to write more freely as well!
This post is therefore dedicated to the Web206 students who were my captive audience yesterday, as I improvised a lecture for them. I hope that anyone visiting this site who has listened to the lecture since (because, yes, it was recorded too) will forgive the conversational section at the end. I tried to make my discussion with the students in the lecture theatre work for the recording, but I don’t think that I was completely successful.
I couldn’t overlook this new robot, erm, pillow-bear.
While its communication skills might seem limited to pawing at your head, this robot listens for your snores while it’s cute companion monitors your blood oxygen. Should your sleep become less than silent, or your oxygen levels drop alarmingly, the pillow teddy will wake you with a gentle paw on your head (although sound sleepers might require a more energetic thump I suppose).
It’s a little difficult to see from the video just how effective the tickling/thumping action might be, and it’s also a pity that the level of background noise makes it hard to relate the bear’s reaction to the snore of the presumably willing test subject (or more likely the creator of the robot).