any time anonymity is involved, people misbehave
Call and response
Mar 8, 2007
The Economist
Computing: Nobody enjoys telephoning a call centre. Could “chatbot” technology make the experience less painful?
With their irritating menu trees and endless holding for the next available operator, call centres are one of the bugbears of modern life. Could a dose of software make the experience of dealing with one less painful? Researchers working on conversational software programs, or chatbots, certainly hope so. They aim to supplement and even replace human operators with software that can understand ordinary conversational language, and thus deal with calls more efficiently.
IBM, for example, has developed a “speech analytics” system that listens in to a call-centre conversation and spots keywords and phrases to help the operator handle the caller's query. One of the limiting factors in dealing with a query is how fast the agent can match the caller's problem to a “knowledge base” of thousands of possible solutions, says David Nahamoo, chief technology officer for speech technology at IBM's T.J. Watson Research Centre in Yorktown Heights, New York. The keywords are used to search the knowledge base and quietly make suggestions to the operator, so speeding things up.
Another system developed at IBM, called Sensei, helps to ensure that the caller and operator can understand each other properly by assessing a potential operator's pronunciation, grammar and comprehension. The system has been used to help select candidates for call-centre positions in India, and to train them once they have got the job, says Dr Nahamoo. By assessing and improving operators' ability to speak a second language, Sensei ensures that operators are easy to understand and can deal with callers efficiently. And although the prospect of being interviewed for a job by a piece of software might sound dehumanising, feedback from candidates has in fact been very encouraging, he says. Both of these systems are part of a more general long-term trend towards automating call centres, says Dr Nahamoo. “More and more of the low-level functions are going to be automated,” he suggests.
Rollo Carpenter, a chatbot programmer, hopes to automate the high-level functions too, by developing software that can directly replace human operators. Chatbots have already been used by some companies to provide customer support online via typed conversations. Their understanding of natural language is somewhat limited, but they can answer basic queries. Mr Carpenter wants to combine the flexibility of chatbots with the voice-driven “interactive voice-response” systems used in many call centres to create a chatbot that can hold spoken conversations with callers, at least within a limited field of expertise such as car insurance.
This is an ambitious goal, but Mr Carpenter has the right credentials: he is the winner of the two most recent Loebner prizes, awarded in an annual competition in which human judges try to distinguish between other humans and chatbots in a series of typed conversations. His chatbot, called Jabberwacky, has been trained by analysing over 10m typed conversations held online with visitors to its website (see jabberwacky.com). But for a chatbot to pass itself off as a human agent, more than ten times this number of conversations will be needed, says Mr Carpenter. And where better to get a large volume of conversations to analyse than from a call centre?
Mr Carpenter is now working with a large Japanese call-centre company to develop a chatbot operator. Initially he is using transcripts of conversations to train his software, but once it is able to handle queries reliably, he plans to add speech-recognition and speech-synthesis systems to handle the input and output. Since call-centre conversations tend to be about very specific subjects, this is a far less daunting task than creating a system able to hold arbitrary conversations.
That said, there is more to handling call-centre queries than simply understanding language and looking things up in databases. Sheryl Brahnam, a researcher at Missouri State University in Springfield, suggests that it will also be necessary to program chatbots to deal with verbal abuse. In some cases, she says, companies that have used chatbots to handle online queries have found that when confronted by verbal abuse or sexual innuendo, the chatbots were programmed to respond inappropriately in kind, with insults of their own.
Dr Brahnam has also found that the appearance of the chatbot's on-screen persona, or avatar, has a significant impact on how much abuse is levelled at it. “My study showed that you get more abuse and sexual comments with a white female compared with a white male,” she says. Black female avatars were the most abused of all. This leads Dr Brahnam to question how effective IBM's electronic-elocution lessons will prove to be. Even if two operators are using the same script, she says, some callers may respond differently (or even abusively) depending on the operator's gender or accent.
Never mind the philosophical question of whether it is wrong to insult a machine. To neutralise such situations, chatbots must be able to handle verbal abuse constructively, says Dr Brahnam. She is now devising ways to program chatbots with the sorts of rules that human operators use. There are two broad approaches. The first is a “three strikes and you're out” approach in which the chatbot repeatedly warns the customer to stop being abusive, and eventually hangs up or passes the call over to a human manager. The second approach is more psychological. Giving some ground to customers and acknowledging that they have been wronged, and that their frustration is legitimate and understandable, can help to restore calm and allow the call to proceed.
Chatbots may well be able to speed up the handling of call-centre enquiries. But the same technology could also make things worse, if it is unable to understand callers or gives inappropriate responses. Dr Nahamoo says he doubts that call centres will ever be completely automated, because there will always be some queries that demand human intervention. “There's a balance between cost and customer experience,” he suggests, somewhat diplomatically.
http://www.economist.com/research/articlesBySubject/displaystory.cfm?subjectid=348909&story_id=E1_RSGGDDN
Mar 8, 2007
The Economist
Computing: Nobody enjoys telephoning a call centre. Could “chatbot” technology make the experience less painful?
With their irritating menu trees and endless holding for the next available operator, call centres are one of the bugbears of modern life. Could a dose of software make the experience of dealing with one less painful? Researchers working on conversational software programs, or chatbots, certainly hope so. They aim to supplement and even replace human operators with software that can understand ordinary conversational language, and thus deal with calls more efficiently.
IBM, for example, has developed a “speech analytics” system that listens in to a call-centre conversation and spots keywords and phrases to help the operator handle the caller's query. One of the limiting factors in dealing with a query is how fast the agent can match the caller's problem to a “knowledge base” of thousands of possible solutions, says David Nahamoo, chief technology officer for speech technology at IBM's T.J. Watson Research Centre in Yorktown Heights, New York. The keywords are used to search the knowledge base and quietly make suggestions to the operator, so speeding things up.
Another system developed at IBM, called Sensei, helps to ensure that the caller and operator can understand each other properly by assessing a potential operator's pronunciation, grammar and comprehension. The system has been used to help select candidates for call-centre positions in India, and to train them once they have got the job, says Dr Nahamoo. By assessing and improving operators' ability to speak a second language, Sensei ensures that operators are easy to understand and can deal with callers efficiently. And although the prospect of being interviewed for a job by a piece of software might sound dehumanising, feedback from candidates has in fact been very encouraging, he says. Both of these systems are part of a more general long-term trend towards automating call centres, says Dr Nahamoo. “More and more of the low-level functions are going to be automated,” he suggests.
Rollo Carpenter, a chatbot programmer, hopes to automate the high-level functions too, by developing software that can directly replace human operators. Chatbots have already been used by some companies to provide customer support online via typed conversations. Their understanding of natural language is somewhat limited, but they can answer basic queries. Mr Carpenter wants to combine the flexibility of chatbots with the voice-driven “interactive voice-response” systems used in many call centres to create a chatbot that can hold spoken conversations with callers, at least within a limited field of expertise such as car insurance.
This is an ambitious goal, but Mr Carpenter has the right credentials: he is the winner of the two most recent Loebner prizes, awarded in an annual competition in which human judges try to distinguish between other humans and chatbots in a series of typed conversations. His chatbot, called Jabberwacky, has been trained by analysing over 10m typed conversations held online with visitors to its website (see jabberwacky.com). But for a chatbot to pass itself off as a human agent, more than ten times this number of conversations will be needed, says Mr Carpenter. And where better to get a large volume of conversations to analyse than from a call centre?
Mr Carpenter is now working with a large Japanese call-centre company to develop a chatbot operator. Initially he is using transcripts of conversations to train his software, but once it is able to handle queries reliably, he plans to add speech-recognition and speech-synthesis systems to handle the input and output. Since call-centre conversations tend to be about very specific subjects, this is a far less daunting task than creating a system able to hold arbitrary conversations.
That said, there is more to handling call-centre queries than simply understanding language and looking things up in databases. Sheryl Brahnam, a researcher at Missouri State University in Springfield, suggests that it will also be necessary to program chatbots to deal with verbal abuse. In some cases, she says, companies that have used chatbots to handle online queries have found that when confronted by verbal abuse or sexual innuendo, the chatbots were programmed to respond inappropriately in kind, with insults of their own.
Dr Brahnam has also found that the appearance of the chatbot's on-screen persona, or avatar, has a significant impact on how much abuse is levelled at it. “My study showed that you get more abuse and sexual comments with a white female compared with a white male,” she says. Black female avatars were the most abused of all. This leads Dr Brahnam to question how effective IBM's electronic-elocution lessons will prove to be. Even if two operators are using the same script, she says, some callers may respond differently (or even abusively) depending on the operator's gender or accent.
Never mind the philosophical question of whether it is wrong to insult a machine. To neutralise such situations, chatbots must be able to handle verbal abuse constructively, says Dr Brahnam. She is now devising ways to program chatbots with the sorts of rules that human operators use. There are two broad approaches. The first is a “three strikes and you're out” approach in which the chatbot repeatedly warns the customer to stop being abusive, and eventually hangs up or passes the call over to a human manager. The second approach is more psychological. Giving some ground to customers and acknowledging that they have been wronged, and that their frustration is legitimate and understandable, can help to restore calm and allow the call to proceed.
Chatbots may well be able to speed up the handling of call-centre enquiries. But the same technology could also make things worse, if it is unable to understand callers or gives inappropriate responses. Dr Nahamoo says he doubts that call centres will ever be completely automated, because there will always be some queries that demand human intervention. “There's a balance between cost and customer experience,” he suggests, somewhat diplomatically.
http://www.economist.com/research/articlesBySubject/displaystory.cfm?subjectid=348909&story_id=E1_RSGGDDN