'Press 1 to Feel Infuriated' - Why computerized "voices" mess with your emotions.

When BMW first introduced an in-car navigation system to customers in Germany, the automaker came up against a surprising problem. The system performed without a hitch, providing accurate map routing and location information—but many drivers complained that they were uncomfortable taking directions from a female voice.

BMW engineers headed back to the drawing board to reconsider just who, exactly, should be telling their drivers where to go and whether their car needed more wiper fluid. A golfing buddy? A servile chauffeur? How about a nagging in-law?

In the end, BMW conceded to the complaining customers (regardless of their unfair stereotyping) and settled on the persona of a co-pilot: a knowledgeable and responsible partner, but one who defers to the pilot. In addition to supporting the driver with its programmed expertise, the system had a tone of voice that approximated a co-pilot’s cool and measured temperament.

Brain speak

BMW’s “gender fender bender” puts the headlights on a neurological quirk. According to Clifford Nass, a professor in the department of communication at Stanford University, our brains rarely make distinctions between human voices and computer voices. Because the brain does not have separate areas for human speech and technology-generated speech, he says, we respond the same way to each. All of the emotion, judgment, and cognition that naturally flows from language are potentially in play whether we’re having a heart-to-heart with a loved one or picking a flick with the Moviefone guy.

“Listeners and talkers cannot suppress their natural responses to speech, regardless of its source,” says Nass in the book Wired for Speech (co-authored by scientist Scott Brave). Nass, who was consulted on the BMW project, elaborates that “these technologies, like the speech of other people, activate all parts of the brain that are associated with social interaction.”

Engineers of voice-based technologies are learning now what neurologists have known for a while: Speech is more than the transfer of words. Inflection, pitch, speed, volume, gender, and ethnicity all set a context for speech that is easily as influential as actual content. The brain processes incoming and outgoing language with so many neurological stopovers to account for these elements that it’s impossible to listen without assigning social qualities to the speaker, whoever—or whatever—it may be.

For emotional companionship, press 2

Asked which commercial products are using voice interfaces effectively today, Mark McClusky, products editor at Wired magazine, responds, “My favorite at the moment is a service called Jott, which allows you to manage reminders and a to-do list over voice recognition. It’s surprisingly accurate, and the key is that it focuses on just a couple of uses rather than being too sprawling.”

Accuracy and a discrete application—this is where speech technology functions most usefully for customers right now. One reason Jott succeeds is that it simply flips speech to text without wading into social territory by laying claims to humanness. In short, it’s not messing with our brains. Granted, refrigerators that talk are more likely to say, “You’re out of butter” than something like “Sometimes I feel so hopeless,” but plenty of services, especially consumer-help lines, do attempt to emulate genuine communication. Some even apologize when they misinterpret a human speaker.

As McClusky notes, “Who hasn’t ended up shouting at a voice system, trying to get it to understand what you’re saying?” We often find ourselves infuriated by computerized voice services not only because they’re incapable of directly addressing a request but because they remove us from the decision making, which is a natural, human element of spoken interaction. All of the options are theirs. It’s the broken promise of such voice interfaces that makes them so frustrating; they want us to respond to them like people, but they treat us like machines.

Driven to safety

We know better than to expect a computer voice to truly know the person it’s addressing, yet our brains have difficulty bridging that gap. Fortunately, technology moves faster than evolution does. Speech-based technology will learn how to cater to human response far sooner than our brains will grow a new node dedicated to computer-generated voices.

Case in point: An experiment cited in Nass’s book describes a safety test wherein the voices on car interfaces were paired with the emotions of drivers. In one group, enthusiastic computer voices were used for happy drivers and subdued voices for upset drivers; in a second group, the emotions of the interfaces and drivers were intentionally mismatched. Even with the same exact words used by the computerized voices, there were less than half as many accidents when the interface matched the emotional state of the driver.

The horror of HAL

One of the best examples of how our brains assign human qualities to computer speech hails from 1968, when speech-recognition technology was the stuff of sci-fi. That year, a computer named HAL 9000 controlled a space voyager in the movie "2001: A Space Odyssey." Once astronaut Dave Bowman realizes the computer is trying to kill him, he goes to HAL’s core to shut it down once and for all. The calm, detached voice of HAL is heard calmly begging for its “life.” And it’s horrifying. How could a voice sound so unperturbed when its memory is being pulled, a computer’s equivalent of a lobotomy?

Stop, Dave ... I’m a-fraid, Dave.... Dave … My mind is going … I can feel it.

Hey, HAL—too bad for you. We can feel it, too.

More Brain & Body Columns on MSN Health & Fitness:

• Why Do Drunk Drivers Reoffend?
• Why Sex Sells and Diets Fail
• Medical Intuition: Wonder Healing or Junk Science?

Search: All Brain & Body Columns

Rich Maloof's award-winning writing has covered subjects ranging from soda pop to stem cells. He has written for MSN, CNN, MSNBC, Yahoo!, Women’s Health, and various other publications. He is the published author of 12 books to date, including several instructional titles for musicians. His latest title, This Will Kill You, is being released by St. Martin's Press in May. Rich is a regular contributor to Brain & Body.