Research Innovation Growth
Innovation Challenge Develop an Artificial Intelligence computational framework to quantify human behavior from recorded videos in the context of job interviews
In effect, build a computational framework that can automatically predict the overall rating of a job interview given the audio-visual recordings and score them on behavioral competencies.
Real world hiring decisions are based on assessments of knowledge and skills as well as holistic judgments of person-job fit. Empirical studies show that non-verbal behavior is as important as the verbal response in job interviews.
The success or failure of the interviewee's effort is traditionally assessed subjectively by the interviewer, either through a holistic impression or quantitative ratings. The validity and reliability of these assessments is subject to much debate.
The new emerging alternative to the traditional human-only interview assessment model is to augment human judgment with automated assessment of interview performance. This is being done by conducting robot video-based interviews bringing many benefits to both interviewers and interviewees, including the convenience of offline reviewing and decision making by human resources enables 4X faster hiring.
The challenge is to build an automated scoring model to predict behavioral traits of candidates who take their robot interviews using their recorded videos.
EvueMe Selection Robot has developed a Robot Interview platform where a human looking Robot interviews the candidates.
Questions to judge the required competencies for a particular job are programmed in our robot. Candidates get an interview invite and call our Robot interviewer "On-Demand" all 24 hours of the day as per their availability.
Once the Robot appears on the screen, she makes the candidate comfortable speaking to her with practice questions and then asks the actual interview questions. The answers to each question by the candidates are recorded as separate videos.
The challenge would involve 3 key steps:
Development of a standardized protocol for collecting structured interview (SI) videos and a set of rubrics for the human rating of the recorded performance.
Development of a model to automatically extract features from the interview videos. These extracted features will come in 3 buckets
With improved human scores and automated multimodal features, build automated scoring systems to predict several behavioral traits like temperament, ambition, communication, cooperative, systematic, integrity, creativity etc. in the job interview
A computational framework that can automatically predict the overall rating of a job interview given the audio-visual recordings and score them on behavioral competencies built by analyzing at least 300 recorded interviews.
