Predicting job interview performance

What makes an interview go well? From 138 mock interviews of MIT undergraduates with professional career counselors, our system predicted how judges rated candidates on traits like excitement, friendliness and engagement, and explained which behaviors mattered.

The MIT mock-interview setup with two cameras.
138mock interview videos (about 10.5 hours) of 69 MIT undergraduates
0.73+correlation with averaged ratings from 9 judges for key traits
Prosodyhow you say it mattered more than words or facial expressions

What you say and how you say it

Facial expressions, language and prosody were extracted automatically. Practical advice emerged: speak more fluently, use fewer filler words, say 'we' more than 'I', use more unique words and smile more. Some feature combinations predicted higher ratings only for male candidates, which led to follow-on work on gender and automated feedback.

  • Automated Prediction of Job Interview Performance: The Role of What You Say and How You Say It I. Naim, M. I. Tanveer, D. Gildea, and E. Hoque IEEE Automatic Face and Gesture Recognition (FG 2015), Slovenia, 2015 · Oral presentation Predicts job-interview ratings from both what candidates say and how they say it.
  • Automated Analysis and Prediction of Job Interview Performance I. Naim, M. I. Tanveer, D. Gildea, and E. Hoque IEEE Transactions on Affective Computing 9, 2018 AI that predicts how interviewers will rate a candidate from what they say and how they say it, and explains which behaviors matter.
  • Analyzing the Impact of Gender on the Automation of Feedback for Public Speaking A. Singhal, M. R. Ali, R. A. Baten, C. Kurumada, E. Marvin, and E. Hoque International Workshop on Human Behavior Understanding (HBU), FG 2018, 2018

← All research