BBC NEWS Americas Africa Europe Middle East South Asia Asia Pacific Arabic Spanish Russian Chinese Welsh

 You are in: Sci/Tech
Front Page 
UK Politics 
Talking Point 
In Depth 

Commonwealth Games 2002

BBC Sport

BBC Weather

Monday, 1 October, 2001, 09:31 GMT 10:31 UK
Robots say it with feeling
'Cynthia' at this robotic bar PA
Robot bartender tells jokes in an American voice
By BBC News Online's Alfred Hermida

People read personality into a synthetic voice even when they know that it comes from a computer, US researchers have found.

They are also more readily influenced when the computer's voice mirrors their personality.

"The synthesiser is perceived as more effective and, more importantly, the web page is seen as smarter and more likeable if the voice sounds more like the personality of the person," Clifford Nass, at Stanford University, told the BBC's Go Digital programme.

The findings could affect the future development of text-to-speech systems, which can be used to read web pages over mobile phones.

Voices with personality

The aim of the experiments at Stanford University was to find out how strongly people attributed emotion, personality, gender and intent to voices that were obviously not human.

It seems to be impossible for the human brain to turn off its attempt to figure out 'who is this person I am speaking to'

Clifford Nass, Stanford University
In one experiment, 72 people listened to book reviews from a mock bookshop on the web. The synthesised voices were made to sound either extroverted or introverted.

The researchers found the participants were more likely to say they would buy the book when the "personality" of the voice reading a book review matched their own.

The results were confirmed by other experiments.

Subconscious reaction

"It seems to be impossible for the human brain to turn off its attempt to figure out 'who is this person I am speaking to'," said Professor Nass.

Clifford Nass Nass
Nass: Human brains evolved to focus on the voice
"Even when we know consciously it's not a person, 200,000 years of human evolution have driven up to decide whenever we hear a voice that I had better know this person and understand its personality."

The researchers found that the response was the same even when participants were continually reminded they were talking to a computerised voice.

"Our brains are exquisitely attuned to voices," the professor said.

"Because of that and combined with the fact that humans are incredibly social animals, it becomes natural for us to leap into the mode of drawing as much information as we can, not just from the language and the words, but also the sound characteristics."

In cases where the computer told a joke, some people joked back, even though they did not expect the computer to get the joke. They simply felt it was the right thing to do.

Comfortable computing

The researchers believe their findings have significant implications, as content providers could use this information in order to make their products more appealing and persuasive.

Most people don't enjoy working with computers. They insult us all the time, they don't worry about our personality or style, or do anything that makes us feel more comfortable

Clifford Nass, Stanford University
"Most people don't enjoy working with computers. One fundamental reason for this is that they make an incredible number of faux pas," said Professor Nass.

"They insult us all the time, they don't worry about our personality or style, or do anything that makes us feel more comfortable.

The more the computers make us feel comfortable, the more they fit in with our social expectations, the better we'll feel and the more effective they will be."

The other key use is in text-to-speech systems. These are becoming more popular because they make computers and internet content more accessible to the visually impaired and blind, as well as to illiterate sighted people.

"Many people want access to the web through a medium other than computers and the telephone is a very well established medium," said the professor.

"To get access to web content via the phone, you need a voice and realistically that has to be a synthesised voice."

But the voice used needs to be appropriate to the content.

Happy or sad

The researchers found people would freak out when they heard a male voice promoting a female product and vice versa. This was also reflected in experiments with happy or depressed sounding voices reading the news.

"People were terribly disturbed when a cheery voice told them about some terrible news story or conversely when a sad voice talked about a happy story.

"It would actually change the tone of the story and also reduce credibly. It would make people not trust this story, think something was wrong."

The research appears in the Journal of Experimental Psychology: Applied, published by the American Psychological Association.

Clifford Nass
"We draw information from language, words and sound characteristics"
Clifford Nass
"Computers insult us all the time"
Synthesised voice
Hear the computer tell a dodgy joke
See also:

21 Sep 01 | Artificial intelligence
Predicting AI's future
12 Sep 01 | Artificial intelligence
Computer babbles like a baby
10 Sep 01 | Artificial intelligence
Past is the future for Hollywood's robots
10 Sep 01 | Artificial intelligence
Timeline: Real robots
27 Sep 01 | Sci/Tech
Acceptable face of robotics
Links to more Sci/Tech stories are at the foot of the page.

E-mail this story to a friend

Links to more Sci/Tech stories