Conversation coach software works like a social simulator
By Ben Coxworth
June 18, 2013
Whether it’s a job interview or a hot date, there are certain interpersonal situations where we really want to be at our best. In some cases, we may even run through possible conversational scenarios in our heads beforehand, in order to “train” for the big event. The problem is, those imaginary interactions can’t provide us with unbiased feedback on what we could stand to improve. MIT’s new MACH (My Automated Conversation coacH) software, however, does exactly that.
MACH is designed to run on an ordinary laptop, and utilizes the computer’s webcam to track the user’s movements and facial expressions, and its microphone to follow their speech. It features a three-dimensional animated face that engages in a simple conversation with the user – not only does that face ask and respond to questions, but it also smiles and nods in response to the user’s speech and motions.
Just as with many human-to-human conversations, however, MACH is actually scrutinizing the user the whole time. As the person talks, the program analyzes factors such as their eye contact, head gestures, frequency of smiles, volume and speed of speech, and use of filler words along the lines of “like” and “basically.”
Once the conversation is complete, the user is provided with a video of their performance, along with a detailed analysis that points out things they need to work on.
In a test of the system, 90 MIT juniors each took part in two mock job interviews with campus career counselors. After the first interviews, one third of the test subjects watched videos featuring interview advice, one third used MACH but received no feedback, and one third used MACH and did receive feedback. They then went for their second interviews, with different counselors who didn’t know about their prepping.
That second batch of counselors rated the subjects’ job-worthiness about the same as the first counselors did, with the exception of the students who had used MACH and received feedback. That group scored significantly higher on their second interviews, when it came to criteria such as “appears excited about the job” and “would you recommend hiring this person?”.
MACH was developed by a team led by MIT Media Lab doctoral student M. Ehsan Hoque, who believes that it could be particularly useful for people with conditions such as social anxiety or Asperger’s syndrome. The software can be seen in use in the video below.