Innovation at EASYRECRUE is a priority. Investing in Research and Development is therefore key to improving our products and having a real impact on the HR world.
This is why we have welcomed to the team PhD student Leo Hemamou and his engineer partner and co-author Ghazi Felhi to conduct research on how to make the best use of pre-recorded video interviews using artificial intelligence. The dynamic duo was selected to present their work at the 8th International Conference on Affective Computing & Intelligent Interaction (ACII) in Cambridge, UK, on September 3rd to 6th. This conference is the premier international forum for research on affective and multimodal human-machine interaction and systems.
Leo and Ghazi were invited to speak about their research focus: developing the workforce through affective computing. In other words, they are presenting a new AI methodology to automatically analyze candidates' reactions during pre-recorded video interviews.
Here is an overview of how they do this and how it will benefit recruiters in the near future.
1# Targeting important moments of the interview
Leo and Ghazi’s research focuses on video interviews and aims to prove that some specific moments in the interview are more relevant in understanding a candidate. These moments are called attention slices.
But how to define what moments are more important?
Here is where artificial intelligence comes in. The computer analyzes the candidate’s facial expressions and in particular the eyes – how closed or opened are they? – the mouth, - is the interviewee pinching her lips (lip stretcher and lip tightener for the experts reading us), and the activation of the jaw drop.
These key moments mostly happen in situations where the candidate is stressed or anxious. Closing one’s eyes, pinching one’s lips or jaw movements are signs of concentration or stress. These indicators echo Deborah Powell's article "Behavioral Expression of Job Interview Anxiety".
Attention slices will happen predominantly at the beginning and at the end of a job interview, when the candidate is asked to speak the most, therefore being more expressive.
2# Drawing conclusions from the analysis of attention slices
Important moments give indicators on a candidate’s hirability. Leo Hemamou conducted a test using other random slices in video interviews to prove this. The results showed a clear difference between these two types of situations in a single video; 3 seconds of an attention slice always contains more valuable information on the candidate than any other random segment of the interview. The computer can collect a meaningful amount of data in a short timeframe (a few seconds) thanks to the number of visual cues given by anxious or nervous candidates.
The purpose of this analysis would be to highlight, through deep learning, these significant moments for recruiters, as a decision-support tool. Additionally, presenting such specific indicators helps develop trust between AI tools and users.
This methodology is also designed to highlight potentially influential social signals that have not been explored by researchers in organizational and industrial psychology.
Furthermore, the results of this research could enhance the candidate's experience by giving her feedback based on the analysis of the important moments of her video, thus helping her improve on her interview performance.
Interested in reading more? Download the poster our researchers presented at the ACII conference!
Also check out our ebook to learn more about how AI can help HR professionals optimize their recruiting process.