Scaling Student Support with Conversational Artificial Intelligence

Of the students who enroll in higher education, 48 percent will not earn a degree within six years (Kena et al, 2015). College enrollees who fail to complete disproportionately come from underserved communities that may lack the resources to help students navigate challenging financial, academic, and social situations. The impacts of leakage from the college-going pipeline reverberate across multiple aspects of society. For example, the high rates of college dropout costs the US an estimated $4.5 billion in lost earnings, which, in turn, reduce annual tax revenues (American Institutes for Research, 2011). We aim to design, implement and test an innovative solution to the issue of college dropout and to ensure that the findings benefit students and universities beyond our site of implementation. To address the related issue of ?summer melt,? whereby college-intending students fail to transition to postsecondary education, in February of 2016, we introduced an artificially intelligent (AI) virtual assistant into Georgia State University?s enrollment process. The virtual assistant: 1) nudges students with reminders relevant to their individual enrollment and matriculation processes, providing them with timely answers to their questions; 2) supplies university staff with greater insight into individual student mindsets; (3) frees time for school administrators to engage deeply with students who most need support; and 4) endows the university with a unique strategy to provide timely, personalized support. Through the pilot with GSU, our experimental study showed that the virtual assistant significantly improved matriculation rates of committed would-be freshmen by 3.3 percentage points (Page & Gehlbach, 2017). This result parallels prior research on summer melt (e.g., Castleman & Page, 2015). However, because the smart virtual assistant technology minimizes staff time needed for straightforward student communication, GSU admissions and financial aid staff could redeploy their time on issues that only experienced counselors can solve, making this intervention extraordinarily promising for scaling up. Building on this initial success, the goal of the proposed project is to extend the use of the virtual assistant (AI) technology to improve college persistence through a pre-registered field experiment. Specifically, we propose to test the technology?s capacity to improve students? academic success and retention at California State University, Northridge (CSUN).