Anthropomorphic Inter faces — So Misunderstood

April 2015
By Bryan Bergeron

I’m writing this in an airport terminal, seated next to one of those talking rear projectors. You’ve no doubt seen or heard one of these in action. The projection of a smartly dressed woman on a cutout screen describes some aspect of security, while the passengers do whatever they can to distance themselves from this noise source as they try to navigate the TSA gauntlet. From what I’ve seen, an ordinary flat screen would have been a better communications vehicle, and much less of an obstacle for foot traffic.

Apparently, the creators of the unit thought that passengers would pay more attention to the human form when projected onto a human-shaped screen. However, people aren’t so easily fooled. While these units may have been curiosities at one time, today the novelty has worn off.

The underlying problem with these talking projectors is that user expectations can’t be met. What’s the point of a huge footprint and anthropomorphic display that is totally passive? A large wall-mounted touchscreen that provides information on demand instead of a loop of recorded audio would provide more value to users. I’ve used these information kiosks and they’re well received. When it comes to connecting with users, it takes more than a pretty face.

It’s the same with robots. Adding a face and skin to a robot to make it look more human-like can backfire unless the capabilities of the robot match the façade. For example, take the series of female androids that have been developed at Osaka University over the past five years. This evolving line of robots look like real women — complete with facial expressions. The latest models can speak, as well as respond to voice questions. In addition, these robots have captured the attention of the general public because they so closely resemble real humans that they are often described as “creepy.” I’d argue that the anthropomorphic interfaces used by these experimental robots — while a valid experimental vehicle — are at best a distraction for practical applications.

Taking a lesson from computer games, computer-generated characters are believable when they’re rendered in either low resolution or ultra-high resolution. Computer-generated characters that are rendered somewhere in between are a distraction. The user spends mental effort trying to figure out what’s wrong with the character instead of focusing on the game. With low resolution, it’s obviously a character and the user moves on. At the other extreme, the user accepts the character as real and also moves on with game play.

Most of what I’ve seen in the way of anthropomorphic interfaces for robots is in the low resolution area — and that’s fine. I’d rather be pleasantly surprised with what a robot can do than be let down by the promise of greater capabilities suggested by an anthropomorphic interface.

It will likely take a decade or more before the insides of a robot can match the expectations of what we can currently present from the outside. No doubt, when we reach that point in the evolution of robotics, we won’t even notice it. Like the smartphone, most of us will just take it for granted. I also know that there will be experimenters out there tearing down and modding those initial models to get even more out of the hardware and software. SV