Steven Z
- Research Program Mentor
PhD at George Mason University
Expertise
Psychology, leadership, organizational behavior, statistics, psychometrics
Bio
Steven Zhou earned his Ph.D. in Industrial & Organizational Psychology program, studying under Dr. Stephen Zaccaro and Dr. Philseok Lee. He previously received a B.A. in Industrial & Organizational Psychology and M.A. in Religion from Pepperdine University. Prior to starting the Ph.D. program, he worked in human resources and data analytics for a 250+ FTE sales strategy department at a $1.44 billion global food subscription company. He has prior research experience in cross-cultural leadership and college student development, teaching experience in Organizational Behavior and K-12 speech & debate, and extensive corporate training experience including a leadership seminar at the Alibaba International HQ. At Mason, Steven's research centers around leadership, personality, and psychometrics. He is currently working on projects in early childhood leadership development, new methods of conceptualizing and measuring leadership behavior, multidimensional forced choice (MFC) methods of measuring personality, faith and work, and text analysis methods of assessing job ads and syllabi. He won the 2021 International Leadership Association's Kenneth Clark Student Research Award for his work on leader behavior profiles. He also has taught undergraduate statistics and is a graduate TA for Mason's online Master's in I-O program, and was the 2021 Outstanding Grad Student Instructor award winner. Finally, he served as the 2021-2022 President of Mason's Graduate and Professional Student Association (GAPSA), and he now serves as a Graduate Assistant in the Office of the Provost for Graduate Education. Outside of Mason, Steven is an Associate Contributor for Young Voices and Director of Finance & Administration for Project SHORT. Steven is interested in pursuing an academic career and is passionate about teaching, research, and policy in higher education.Project ideas
Developing a new method of measuring multiraciality
Have you ever wondered what happens to data collected from participants who check the box "Multi-racial" when asked about their race or ethnicity? In most cases, those data points are discarded, because we currently lack a reliable and effective method of analyzing racial-ethnic differences among multiracial individuals. This project proposes a new statistical analysis method that potentially allows us to find unique and groundbreaking insights into the collection and measurement of multiracial data.
Identifying gender and race/ethnicity biases in sentiment analysis
Sentiment analysis is a popular form of text analysis that uses machine learning to predict the specific emotion that the writer of the text was feeling. There are several popular machine learning models that can take a chunk of text and "assign" scores on various emotions to it. However, we know from prior research in linguistics that there are differences between gender and racial subgroups on how language is used. Thus, these popular sentiment analysis tools, which do not differentiate between gender and race, might lead to biased scores that have important future consequences. This project investigates the degree to which popular sentiment analysis models exhibit gender and racial-ethnic biases.
Using machine learning algorithms to predict volunteer turnover
Turnover, or employees quitting an organization, is a key issue for most employers, and much prior research has examined the various job-related predictors of turnover. However, few have explored the effect of non-cognitive predictors (e.g., role ambiguity, satisfaction with colleagues) on turnover, especially among volunteer populations. Such research is particularly important with volunteers, who generally lack traditional motivators such as pay or lack of alternative options. Moreover, few studies have employed modern machine learning methods to examine the relative importance of different predictors. This study tests 14 different machine learning models on a large sample of volunteers, using a set of non-cognitive variables to predict turnover.