Researchers at Massachusetts Institute of Technology (MIT) in the US have produced a computer program that helps individuals practise interpersonal skills so they feel more comfortable in job interviews or going on a first date. The individual interacts with an on-screen computer-generated face, while the program monitors and gives feedback on their body language, eye contact, pace and speech.
MIT Media Lab doctoral student M. Ehsan Hoque led the team that produced the program, which is called MACH (short for My Automated Conversation coacH). He and his colleagues documented how they developed and tested MACH in a paper to be presented at the 2013 International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) in Zurich, Switzerland, in September.
“Interpersonal skills are the key to being successful at work and at home,” Hoque says in an MIT press release.
“How we appear and how we convey our feelings to others define us. But there isn’t much help out there to improve on that segment of interaction,” he adds.
The team believes MACH can help lots of people, particularly those with social phobia, a condition that, according to the National Institute of Mental Health, affects some 15 million adults in the US.
Plus, surveys show that many people have a fear of public speaking, and there are also those, such as people with Asperger’s syndrome, who struggle with maitaining eye contact or responding to social cues.
But while training can help, that in itself can be stressful because it usually involves exposing one’s shortcomings to others. And there are other disadvantages.
Hoque says many people with social phobia would prefer “some kind of automated system so that they can practice social interactions in their own environment”. They want to “control the pace of the interaction, practice as many times as they wish, and own their data”.
Apparently, these are all problems that MACH overcomes. And the researchers have done tests to prove it.
They tested MACH on 90 MIT student volunteers, all native English speakers, whom they randomly assigned to one of three groups.
All three groups went through two simulated job interviews, one week apart. The interviewers were MIT career counselors: a different team of counselors carried out the second interview to the team that did the first one.
After the first interview, all three groups of interviewees received help of some sort. One group watched films giving advice on effective interviewing, another group had a session with MACH but received no feedback on their performance, while the third group use MACH and then watched video recordings of themselves accompanied by a comprehensive analysis of their behavior.
The behavior analysis told the volunteers in the third group how much they smiled, maintained eye contact, varied the tone of their voice, and how often they used “filler” words such as “umm”, or “basically” or “like”.
After both interviews, the counselors were asked to rate the interviewees on various measures such as “overall performance”, “appears excited about the job”, and also to say whether they would recommend hiring the interviewee.
When the researchers analyzed the counselor ratings they found a statistically signficant improvement between the first and second interview for the group that used MACH for practice and feedback, and no signficant change for the other two groups.
The MACH software runs on a laptop. Users address an automated interviewer, a life-size, three-dimensional artificial but realistic face that smiles, nods, and responds to users’ speech and body language. It also asks questions and gives responses.
While the interaction is going on, MACH monitors the user’s face and body via the computer’s webcam, and captures his or her speech through the microphone. The program analyzes smiles, head gestures, volume and speed of speech, use of filler words, and other things.
It may seem odd to use computers to teach people how to interact more effectively with each other.
But apparently such programs can often offer users wanting to learn such skills more than human role-players can, says Jonathan Gratch of the University of Southern California, where he is a research associate professor of computer science and psychology.
“They can faithfully embody a specific theory of pedagogy, and thus can be more consistent than human role-players,” says Gratch, who was not involved in the research behind MACH.
For example, there is evidence that people on the autism spectrum are more likely to improve their understanding of the subtleties of human interaction and social situations when the training targets specific skills such as modeling and role play, note the researchers in their paper.
1 in 50 American schoolkids has autism, according to the US Centers for Disease Control and Prevention (CDC).
Another possible explanation for MACH’s effectiveness could be the feedback doesn’t come from a human: sometimes it is easier to deliver (and accept, perhaps) the brutal, objective truth from a machine.
Hoque, who is in the final stage of completing a doctorate in media arts and sciences, says they initially designed MACH to help people with job interviews, but there is no reason why it couldn’t also train them for other kinds of social settings.
In a bid to “help the world and improve lives” the German software company SAP said recently it plans to hire hundreds of people diagnosed with autism, because they they want people who “spark innovation” and “think differently”.
Written by Catharine Paddock PhD