“Wizard of Oz” for the Modern Roboticist
Twenty years before the word “robot” entered the modern vernacular, L. Frank Baum wrote his novel, The Wonderful Wizard of Oz, with a robot as a secondary character. I am talking about, naturally, the Tin Man. As a robotics researcher, I cannot help but feel some pang of envy, whenever I read about a fictional robot that moves, talks, and emotes like a human. Numerous other creators, from Isaac Asimov to George Lucas, have envisioned societies with varying degrees of robot integration and companionship. Yet, imagination has raised a question that only science can rigorously answer: how would people interact with robots? Given the limited interactive abilities of most modern robots, some may consider this a silly question, akin to asking how people would interact with a toaster. Others may consider it a hasty question, one meant for a human society in the year 2100 or afterward. I believe that this question is neither silly nor hasty, and that we can produce meaningful answers before something like the Tin Man becomes reality.
At the 2019 Northeast Robotics Colloquium, a forum for cutting-edge robotics research in academia, Professor Wendy Ju (Cornell Tech) outlined her research on a “puppet-master model” for robotics. In one experiment, research subjects sat in a room, reading a book or browsing their phones. A small black ottoman began to move around, positioning itself next to the subject’s legs. Some subjects used the ottoman. Others rejected it and tried to ignore it. One even treated it as a pet, patting it affectionately rather than using it as an ottoman. Ju never told any subject that the ottoman would behave in this manner, but all of them deduced its “intentions” somehow.
By all appearances, the ottoman was a robot, fully automated down to its affectionate, almost cute, emotional tics. But this was not so. You may have heard of the “mechanical Turk”, a chess-playing automaton from the early 1800s. It was later found to be a fraud, controlled by a human inside the machine. Like the mechanical Turk, the mechanical ottoman was operated (remotely) by a human. This is because all but the simplest behaviors of an autonomous robotic ottoman are out of reach to modern knowledge. The key idea is robotic control. Ju’s remote-controlled model of a robotic system dispenses with nearly all the features of a truly autonomous robot, but her robot behaves as if it were autonomous. As a result, her research yields useful and interesting insights on the study of human-robot interaction, answering fundamental questions about the nature of robots in our society.
Like the Wizard of Oz (a reference she specifically noted in her talk at NERC), Ju uses simple machinery under the guise of something greater, in order to achieve what the simple machine itself could not. Her methods generalize to other projects. In a separate study, she found that when a robot’s behavior is ambiguous, a user will come to develop their own strategy to control a robot. Additionally, a robot’s behavioral tics and display of emotions (categorized as “sad” or “happy”) triggered empathetic responses in the user. Perhaps most interestingly, if the robot was meant to help humans solve a task, users tended to take more initiative, and many felt annoyed when a robot appeared too proactive.
More generally, Ju noted her thoughts on the notion of ubiquitous robotics, the idea that personal robots may become as commonplace as the smartphone or laptop computer. What if everyone had a “Tin Man”? What if, more realistically, everyone had a few robots for various household tasks? The motivation of her research is to understand how humans behave around robots, well before such robots enter our society. Initially, the technical minds in robotics get a chance to rapidly prototype, deploy, and test ideas about robotic design. But not long after, people as a whole will encounter Version 1.0 of a fully-autonomous home robot that is precisely tailored to their needs. If we wait for autonomous robots before collecting this data, we risk losing a head start of several decades. And in the process, we might just make discoveries about ourselves.
Check out Professor Wendy Ju’s personal website: http://wendyju.com/. Among the many papers she has written, I highly recommend reading “Empathy: Interactions with Emotive Robotic Drawers” (2014) and “Mechanical Ottoman: How Robotic Furniture Offers and Withdraws Support” (2015).