Robot Interviewers

A few years ago I conducted telephone interviews for a business-to-business market research project.  I called librarians to ask their opinions of a proposed new software product for libraries.  One woman told me very politely after I read the introduction that it sounded like an interesting topic to discuss and that she would not mind taking the time, but that she had recently agreed to do a telephone survey that turned out to be an obscene phone call.  She said that she was very sorry, but that she would have to decline.  I had been trained to overcome objections, but this was a new one for me.  I had to mark the response for that library as refused to participate and move on.

The very next day, on the same project, another woman told me that the survey I had called her about sounded interesting, but that she had recently agreed to do a survey that turned into an obscene phone call.  This time I was quicker on the uptake.  I asked the woman if we could have a female interviewer call her.  She told me “Well, since you asked that question, go ahead.”  I was able to complete the interview with her.  I guess some obscene callers have a thing for librarians.

This is one of many anecdotes that I need to keep in mind as I investigate the use of speech recognition technology to collect information for market research and public opinion research telephone surveys.  My motivation for this investigation is not to reduce payroll costs, although that is a consideration.  Nor is my motivation to eliminate the headaches involved in supervising human interviewers who either do not want to work or who do not pay attention to instructions.

My motivation for investigating the use of speech recognition technology for telephone interviewing is that I think robot interviewers could get more accurate information for our clients.  Some of the vendors of systems that use speech recognition technology for market research interviewing use this as a selling point.  They say that it is actually an advantage for a respondent to know that they are being interviewed by a robot because the respondent is more likely to give an honest opinion. 

I am thinking specifically about a survey we did earlier this year.  We called registered voters throughout Ohio to ask them their opinions of proposed legislation regarding animals.  The survey had questions about regulating the ownership of exotic animals, the treatment of chickens on factory farms, and whether the penalty for cockfighting should be a felony instead of a misdemeanor.  I wondered at the time whether people would give honest answers to these questions or if they would give what they considered to be socially acceptable answers.  Who is going to say that they are in favor of cockfighting?  A person might say that if they knew that their answers would be kept confidential and if they did not have to say it to a human interviewer.

We will most likely not use speech recognition technology for telephone surveys until the technology advances quite significantly.  The technology can now be used for simple surveys that have yes/no or agree/disagree questions.  It can skip a question based on an answer to a question if appropriate.  It can record answers to open-ended questions, but cannot probe those responses.  A robot interviewer would not know when and when not to ask “Why?”  The speech recognition systems used in customer service applications can understand what a customer is saying well enough to route a call to a human CSR and can even schedule a reservation, but cannot actually help a customer resolve a billing discrepancy.

Another consideration as I investigate this issue is that researchers and their clients may rely less on telephone polling in the future.  Google claims that its Google Surveys can provide data that is statistically representative of a population.  If so, this will remove an obstacle to doing more research via the Internet.

These are my impressions of speech recognition technology so far.  I would like to find out if I am incorrect.  Future posts on this blog will cover what I learn about speech recognition technology and artificial intelligence as well as successful interviewing techniques as they apply to concepts in market research and public opinion research.  I will also be looking at how big  data can beused for market research and public opinion research and more generally about how automation replaces human workers.  Please feel free to direct me to sources of information or share your own stories about interviewing.

John C. Stevens
jstevens@sapersteinassociates.com