We are preparing a new challenge on the theme of "Speed Interviews".
Companies are already starting to use video interviews as part of the job screening process.
We spend more than half of our waking hours at work. Work conditions our material welfare, our personal development, and happiness. Finding the right job may influence the course of your life for better or worse. This is why it is so important to invest into determining your aspirations, your aptitudes, and material needs. If you are an employer, you a have complementary view of the problem: your hiring choices will condition the material success of your enterprise. To fulfill your specific requirements, you need to attract the right people by making your job offer appealing, specific, and fairly compensated.
The classical methods of recruitment using word-of-mouth, news-paper adds, and classical lengthly on-on-one interviews are ill equipped to face the new employment challenges, which will require to match millions of people with millions of offerings. With informatics, we can represent millions of people and jobs numerically, analyze patterns of good and bad matches, and learn from them using machine learning programs to propose new candidate matches. With new methods developed in research to analyze video data, speech data, handwriting, and natural language, we can greatly enhance the capabilities of current recommendation systems solely based on classical text resumes and questionnaires, getting closer to real interviews.
In 2016, we conducted a first challenge on the theme of FIRST IMPRESSIONS.
According to psychologists and human resource professionals, it takes only a few seconds to people to decide about personality traits of someone they encounter. Can an automatic system do the same? This would be useful both to coach people preparing interviews to make a good first impression and to professionals making hiring decisions to identify what are possible cultural, racial, gender, or other biases they have to fight against to make better, more objective decisions.
We have collected sample video from YouTube of 15 seconds in length.
The personality traits we studied are the "Big Five":
Other possibilities include scoring Competence and Warmth only. We want a trade-off between too many traits (the human scoring the videos will get tired, it will require too much concentration) and too few traits (we will not capture essential information.We need to examine scenarios in which decisions (e.g. hiring decisions or dating decisions) might be based on such criteria. Other considerations might include the selection of "actionable" criteria (criteria that the subject can improve on in a coaching scenario).
We annotated a large number 10,000 15 sec. clips cut out of YouTube videos using AMT workers.