Bizarro bicycle

“Hello there. I’m a robot. Can I help you?”

One of the great Turing moments in the modern world is when you have to deal with a voice-mail system that has been programmed to “behave” like a real person. Depending on how good the voice-acting is and how seamlessly the voice prompts queue up, it’s sometimes possible at the outset of the call to fool people into thinking there’s a real person on the other side. But most people these days are so good at detecting faux-humanity that voice-recognition companies have had to adapt to our enhanced Turing abilities.

David Holsinger, a linguistics expert, has written a fun blog entry describing the pendulum swing that’s taken place in recent years. Voice recognition companies found that two things were happening:

(1) people were so completely fooled into thinking that they were talking to a human that they responded in ways no speech-rec application could ever be built to handle (try matching an utterance like, “Well, howdy! Sure thing — I’m just trying to figure out how many minutes I got left on my phone!”)

(2) the more common response — no one was fooled, and people reacted with equally unparseable globs of “command words,” and frequently words clearly influenced by the predominant visual metaphors in computing: “Clear!” “Restart!”

Obviously, no human interaction has ever included the phrase “Ok, your order comes to $21.99” followed by “No! CLEAR!” Realizing this, voice user interface designers seem now to be dragged by a slow and painful pendulum swing back in the other direction, towards the converational style of “To clear your last message, press STAR-STAR.” … The curtain is up — our users always knew that there was a machine back there, and dammit, they want to treat it like any good machine — by yelling at it in monosyllables.

The interesting thing is that ‘bot manufacturers tend to assume there are only two different type of service ‘bots: i) Ones that can fake you into thinking they’re real, and ii) ones that fail — and are too easily detected as human.

I think there’s another category: iii) The robots that where you know they aren’t human, but don’t care, because they’re robotic in an interesting way. To me, the best robots are like R2D2. They do not try to be human; indeed, they’re explicitly robot-like, and by being robot-like, are actually far more charming than faux-human ones. A good physical example of this is the Roomba: It’s not human-like at all, but because it looks so funny and self-aware as it zips around the room, one tends to treat it as “alive”.

In the voice-prompt world, there’s Wildfire, a voice-activated personal assistant. It doesn’t seem particularly lifelike, but the acting was done by a woman with such a gorgeously husky voice that several male executives I knew used to sit around asking her the time over and over again, to enjoy the sound of her whispering in their ears.

blog comments powered by Disqus

Search This Site


I'm Clive Thompson, the author of Smarter Than You Think: How Technology is Changing Our Minds for the Better (Penguin Press). You can order the book now at Amazon, Barnes and Noble, Powells, Indiebound, or through your local bookstore! I'm also a contributing writer for the New York Times Magazine and a columnist for Wired magazine. Email is here or ping me via the antiquated form of AOL IM (pomeranian99).

More of Me


Recent Comments

Collision Detection: A Blog by Clive Thompson