The First-Year Interest Group (FIG) program at the University of Washington (UW) is a variant of University First Year Seminars (FYSs), which are found at nearly all accredited four-year US colleges and universities. FYSs are believed to boost social cohesion, support knowledge acquisition, and introduce first-year students to the breadth of academic choices available to them at the host institution. At the UW, the FIG program began in the early 1980s, and has grown to involve roughly half of all entering freshmen. Despite the widespread implementation of FYSs at universities, little research has been done to measure the efficacy of these programs. This study leverages UW registrar data (including transcript records and demographic information) on over 60,000 UW freshmen across 13 years to examine the performance of FIG and non-FIG students (as measured by graduation, retention, and grades) through econometrics-based analyses of transcript records. To supplement this quantitative analysis, we use text-mining/NLP on 6 years of FIG exit survey data consisting of the responses of 14,000 students to questions regarding their FIG experiences. Our objective in our work is to use UW student information to conduct a thorough audit of the FIG experience and its effects on first-year student performance and retention. We hypothesize that students who take FYSs have a unique academic experience that is reflected in their academic outcomes as well as social integration to campus. We hope our results will motivate further discussion on current practices around FYS both at UW and across the country, while providing quantitative evidence of their impact on students' academic progression.

The U.S is projected to face a shortage of a million graduates in STEM fields by 2022 according to the President’s Council of Advisers on Science and Technology. Much of this is the result of the current attrition rate from university STEM programs. Only half of all students entering the university intending to major in STEM fields graduate with a STEM degree. Decreasing this attrition rate by just 20% (or by 10 percentage points from 50% to 40%) will address over three quarters of this million-individual STEM gap. Therefore, it is of great interest to study the systemic factors that influence STEM attrition. This study will contribute to the existing body of knowledge on STEM attrition by studying the trajectory of students into and out of STEM fields by examining transcript records for every undergraduate student at the University of Washington from 1998 to 2007. As a first step, we will build a model that intuits students' intended majors by calculating affinity scores based on prerequisite classes taken by students. Examining changes in these affinity scores over time will allow for the identification of inflection points where a student's interests are likely changing. With this information, we can identify when students change fields of study and, on a broader level, also identify when they transition away from STEM fields. We hope to then use a supervised machine learning model to identify the features of students' transcript records that predict if and when this attrition will occur. Ultimately, we hope this information will help inform institutional strategy to improve STEM curriculum design, resource allocation, and retention strategies.