National Geographic News: NATIONALGEOGRAPHIC.COM/NEWS
 

 

"Jerk-O-Meter" Measures Phone Rudeness

Brian Handwerk
for National Geographic News
September 7, 2005
 
People know when they're on the phone with an inattentive jerk, but they might not realize how they sound to others. A new telephone technology, dubbed the Jerk-O-Meter, could help.

Massachusetts Institute of Technology media lab researchers have created a device that analyzes psychological cues in the human voice to rate a speaker's interest in the conversation they're having.

The machine connects to a cell phone and picks up on vocal cues, such as how quickly someone speaks, the amount they interrupt, and whether they use other conversational signals such as repeated "yeahs." The cues are used to measure a speaker's engagement on a scale from 0 to 100.

During calls the device gives its owner messages about his or her performance, displaying notes such as "Don't be a jerk!," "Be a little nicer now," or "Wow, you're a smooth talker."

Anmol Madan, an MIT master's degree student, led the Jerk-O-Meter design with MIT professor Alex "Sandy" Pentland.

"I'm the guy who's not paying attention, who's typing in a keyword or doing several other things, while on the phone," Madan reported during a telephone interview. "So I could relate to it."

Madan's team hopes that the invention can help people improve conversations and relationships, both personal and professional, with those on the other end of the line.

"I think they are sort of tapping into the … fact that the enormous level of frustration that we have nowadays is coming out in our voices," said psychology professor Brian MacWhinney of Carnegie Mellon University in Pittsburgh. "So anything that would help would be perceived as great."

Machine Takes Our Measure

The current prototype of the Jerk-O-Meter monitors several key aspects of phone conversations.

"It measures activity levels, or how often you speak," Madan said. "It also uses mathematical logarithms to measure known stresses and deviations of pitch."

Future versions may become even more precise. They may measure factors such as mirroring, when one speaker shows empathy by adopting the other's voice patterns, and engagement, which monitors how much one person influences the other as they take turns talking.

"The fundamental idea is that by just listening to the human voice you can start predicting interest and engagement with about 75 to 80 percent accuracy," Madan said.

Cognitive psychologists are well aware of such valuable cues.

"I'd say that there are definitely individual linguistic markers that carry across a conversation," said Carnegie Mellon's MacWhinney.

"There are things that will come out very quickly. But how well machines are able to pick that up, I just don't know."

The MIT group put its machine to the test by trying to predict the outcomes of some 200 three-minute conversations. They paired off volunteers and randomly assigned them conversational topics.

Afterwards speakers were asked to rate their level of interest from one to ten, and the Jerk-O-Meter was able to anticipate their interest from 75 to 80 percent of the time.

Big Brother Isn't Listening

The machine is hardly one size fits all, however, and significant hurdles may remain. Simply put, not all of us speak or express emotions in the same way.

Dominic Massaro works extensively on speech recognition at the University of California at Santa Cruz.

Massaro finds the Jerk-O-Meter intriguing but notes that the devil is in the details.

"It's a big challenge to recognize speech correctly, and recognizing emotion is also a challenge," he said. "To recognize the basic emotions like happiness, anger, fear, and disgust, the face is more informative than the voice. The voice by itself can be pretty ambiguous."

"We've done studies where we find cues in the voice, like the overall pitch and the amplitude, but they are not definite," Massaro added. "It's a challenge to recognize what emotion is for both humans and machines."

Future versions of the Jerk-O-Meter could become more effective if software allows the machine to learn more about the speaker through constant feedback. The device would then become more accurate with each conversation.

Habitual jerks and others need have no fear, Madan said. He stresses that the device is intended only for use by its owner and not as a means for rating unwitting speakers on the other end of the line.

But in the realm of self-improvement it could be a valuable tool.

"Lot of people have trouble getting feedback about themselves," Carnegie Mellon's MacWhinney said. "It's a resource problem. … [Y]ou have to listen to another person and think of your responses, so it's rather difficult to monitor yourself."

MacWhinney also wonders if we'll like what the machine has to say.

"Seldom have we ever had computers making personality judgments about us," he said. "In some cases they do tell us, You made a mistake, and we're willing to accept that kind of feedback. But people may become a little miffed if it seems to become too invasive."

Free E-Mail News Updates
Sign up for our Inside National Geographic newsletter. Every two weeks we'll send you our top stories and pictures (see sample).

 

© 1996-2008 National Geographic Society. All rights reserved.