Log in   •   Sign up   •   Subscribe  feed icon

Robotic Behavioral Tics Are The Key To Bridging The Uncanny Valley

Even in the face of all the advances we've made in the field of robotics, one thing is abundantly clear: we've a long way to go before we're able to make a robot that truly appears human (or even humanlike). Right now, far too many robots are stuck deep in the uncanny valley. They're almost human, but there's still something off about them, some unusual means of movement or communication which for many makes them downright terrifying. 

One researcher at Plymouth University in the UK believes he may have discovered a means of making robots less intimidating - and far more humanlike. The key, he explains, lies in giving robots little behavioral tics, small patterns and motions which - though they serve absolutely no practical purpose - will make them feel significantly less artificial to human beings.

Such tweaks, believes Read, could be the key to bringing robots out of the uncanny valley. A robot, for example, might chirp and beep like R2-D2, or fidget when asked a particularly difficult question. 

To test their theory, Read and colleague Tony Belpaeme created two separate sounds: a chirpy, cheerful sounding bleep and a terrible, melancholic whine. They proceeded to record footage of Aldebaran Robotics' Nao making each sound after being slapped, kissed, stroked, or having its eyes covered. That footage was shown to three hundred people on the crowdsourcing website CrowdFlower; each participant was asked to rate how they perceived the robots feelings after each action. 

The pair found that though the subjects responded similarly to each of the two sounds, they were far more engaged when the robot made noise than they were when it didn't. 

 "It is enough to choose or generate a random sound," Read explained, continuing that a sound is all that's necessary to inform someone that something important's going on. "It seems to be an easy way to provide rich expression for robots," he added.Read presented his findings at the Human-Robot Interaction conference in Bielefeld, German on March 3. 

Read isn't the only one looking into the ways in which behavioral quirks can give a robot personality. Sean Andrist and his team at the University of Wisconsin-Madison have devised a means of introducing small, random movements into a robot's head rotation motor. As a result, instead of appearing stationary, a robot's head will twitch every now and then.

Of course, that on its own might be more disturbing than welcoming. The team has also introduced a face-tracking camera that allows the robot to look directly at the person it's interacting with, programming in a tendency for it to avert its gaze from time to time; this mimics the human habit of glancing to the side when thinking of how to answer a question. According to a survey of 30 students, this programmed gaze aversion makes a robot appear more purposeful and thoughtful; Andrist also added that the glances also led to fewer "inappropriate interruptions" in conversation between student and robot.

How and where a robot directs its gaze is of vital importance, as well. Ajung Moon from the University of British Columbia discovered that in order for people to feel comfortable with being handed something by a robot, they must lock eyes with it first. The robot must then look at the point in space where it plans to make the transfer. Again, it's all about accurately mimicing the little tics and movements we as human beings have a tendency to make. 

Last, but certainly not least, Anca Dragan of Carnegie Mellon has introduced a small delay to human-robot interactions, mimicing natural human hesitation.

"It may seem less efficient, but people take to these small differences in robot behaviour," explained Dragan. "Such handover motions have to match human expectations. You don't want the robot moving in ways that surprise or shock people."

At the end of the day, that may be one of the biggest problems with robot-human interaction: the 'robot' side of things is just too smooth, too efficient, and too perfect in how it moves and behaves. Introducing small imperfections and tics to a robot's behavior may seem inefficient and wasteful on the surface, but if it makes people more willing to interact with their mechanical counterparts, it seems a small price to pay.