The goal of the study is to use machine learning/AI to quantify more accurate algorithms for scoring interviews, the goal is to develop predictive algorithms for scoring interviews for future success.
Research Interests
Organizations are rapidly adopting machine learning and artificial intelligence approaches to personnel selection and staffing. In some cases, firms are applying artificial intelligence tools as part of their screening process to assess short videos of candidates, before any human being views the applicant. With or without a human interviewer/assessor, can machine learning and artificial intelligence approaches match or exceed the assessment capabilities of human beings?
Calls for interdisciplinary research in the domain of personnel selection and staffing including new testing and assessment methods; reliability and validity of tools; interviewing, detecting, and remediating adverse impacts including bias; and interpretability of algorithms highlight the need for strong empirical studies.
In addition to questions about efficacy, numerous ethical questions concerning the use of machine learning and artificial intelligence have arisen in this domain, where practice is ahead of science. Are these approaches scientifically and ethically sound? What are the effects on people and organizations? How can these tools be used virtuously and justly? What are the risks, and how can they be mitigated?
This project addresses the enormous opportunity to improve the body of knowledge in industrial and organizational psychology, organizational behavior, and human resources management, and brings together a team of researchers suited to the task. The project will draw from a large and detailed repository of candidates and recorded interviews including audio, visual, and transcription of real-world job candidates whose interview and employment outcomes and other characteristics are known.
Funding will be used to support a PhD candidate or postdoctoral fellow, hire and train student research assistants, and meet expenses including hiring of outside expert interviewers/assessors and investment in technical tools, equipment, or computing needs.
Desired outcomes
· Using machine learning and artificial intelligence approaches, quantify and develop more accurate algorithms including:
o lexical features regarding interview content, speaking fluency, and articulation
o prosody, or speaking style, including pitch and vocal intensities, in relation to answers to interview questions and use of expressive language
o facial features and expressions, including hand and head gestures, posture, gazes, and overall visual motion. The computational models should include data on features such as candidate ethnicity, age, gender, degree of physical attractiveness, etc., which can help to reduce bias.
· Compare and evaluate machine learning and artificial intelligence methods vs. more traditional (and often more expensive) methods like manual qualitative coding and expert assessments of interviews.
· Explore and address key ethical/philosophical questions about the development and use of machine learning and artificial intelligence methods in personnel selection and staffing.
· The ultimate intent of the study is to develop and evaluate predictive algorithms for interview settings and explore the ethics around these approaches, using the results to prepare firms, interviewers, and candidates to be more effective while reducing the effects of bias and non-job-related factors in hiring decisions.
Student engagement
Student engagement is critical to success.
· At least half of the funding will be used to support a PhD candidate or post-doctoral fellow in this domain who will collect and analyze data, help develop tools, and contribute to theory and insights with intent to publish results with the faculty researchers in leading journals.
· Training research assistants as human reviewers and assessors for comparison to machine learning and artificial intelligence approaches or potentially to conduct additional data collection, testing, or analysis.