How rude! Jerk-O-Meter rates quality of phone chats

• Article
• Comments ()
• Click-2-Listen
• Videos

By

CAMBRIDGE, Mass. -- Ever wonder whether that spouse, friend or co-worker on the other end of the phone is really paying attention? The "Jerk-O-Meter" may hold the answer.

Researchers at the Massachusetts Institute of Technology are developing software for cell phones that would analyze speech patterns and voice tones to rate people -- on a scale of 0 to 100 percent -- on how engaged they are in a conversation.

Anmol Madan, who led the project while he pursued a master's degree at MIT, sees the Jerk-O-Meter as a tool for improving relationships, not ending them. Or it might assist telephone sales and marketing efforts.

"Think of a situation where you could actually prevent an argument," he said. "Just having this device can make people more attentive because they know they're being monitored."

The program, which Mr. Madan said is nearing completion, uses algorithms to measure levels of stress and empathy in a person's voice. It also keeps track of how often someone is speaking.

"It's an academically proven thing," Mr. Madan said of the math behind those measurements. "There are a bunch of academic papers published about this."

For now, the Jerk-O-Meter is set up to monitor the user's end of the conversation. If his or her attention is straying, a message pops up on the phone that warns, "Don't be a jerk!" or "Be a little nicer now." A score closer to 100 percent would prompt, "Wow, you're a smooth talker."

However, the Jerk-O-Meter also could be set up to test the voice on the other end of the line. Then it could send the tester such reports as: "This person is acting like a jerk. Do you want to hang up?"

To test the program, Mr. Madan and his MIT colleagues recruited 10 men and 10 women -- all strangers to one another -- and brought them into the lab. The researchers paired off the test subjects, with men talking only to men and women talking only to women, and monitored 200 three-minute conversations about randomly selected topics.

After each conversation, the subjects were asked to rate their level of interest on a scale of one to 10. By measuring the speaking style each person had used in the conversation, Mr. Madan was able to predict what score they would give about 80 percent of the time.