The Stated and Hidden Expectations: Applying Natural Language Processing Techniques to Understand Postdoctoral Job Postings

Postdoctoral positions (colloquially referred to as a "postdoc") are a common avenue for engineering and science students to prepare for competitive academic careers in STEM fields upon completion of their Ph.D. Effective postdoctoral learning experiences are facilitated through mentorship to help scholars develop deeper competencies and skills for the professoriate. While the educational and professional gains are likely significant, postdoc scholars remain forgotten in the engineering education community as no rigorous studies on postdoc education experience have been published in top engineering education journals within the last ten years. Moreover, a major component of the postdoc experience is finding a faculty member or research group that will offer mentorship. Another important consideration is that women and people of color tend to be underrepresented as faculty in engineering and computer science (CS), resulting in limited opportunities for graduates to find the same- and/or cross-gender and race mentors. Presently, there is little known about how impressions of a postdoc position or certain mentors may influence the decision to recruit postdocs. The purpose of this paper is to explore and to assess what are the stated and hidden expectations about postdocs during the recruitment process, and to examine the prevalence of gendered language in job postings which may influence aspiring postdocs. We apply the Knowledge, Skills, and Abilities (KSA) as the theoretical framework to answer the following research questions:1) What are the stated expectations for knowledge, skills, and attributes in the postdoc job postings required for engineering and CS postdocs?; and 2) Are there hidden expectations for gender perceptions from postdoc job postings? In this work, we answer these questions using natural language processing techniques with Python. We collect data for engineering and CS postdoc job postings from the publicly available web pages. By utilizing content analysis on this textual data, this paper presents the stated expectations for postdocs by identifying the most frequently required knowledge, skills, and attributes from job postings. By applying linguistic gender-coding methods on the job postings, we were able to categorize masculine-coded job postings and feminine-coded job postings. The results demonstrate that the majority of the postdoc job postings tend to use masculine language, which only further reinforces gender disparities in engineering and CS. To broaden participation, and to create a more inclusive environment for postdocs, it is important that academia reconsider the language used in such postings. The findings from this work are intended to call attention to the gendered language that could discourage underrepresented populations and to provide a foundation for hiring engineering and CS postdocs.