Improving Financial Aid Application Rates Through AI: Evidence from Washington State’s OtterBot Evaluation

• In Event: Financial Aid

Abstract

Washington State has one of the lowest rates of financial aid application completion in the country—a persistent barrier to postsecondary enrollment, particularly for low-income and first-generation students. In response, the Washington Student Achievement Council (WSAC) partnered with MDRC and Mainstay to optimize OtterBot, an AI-enabled text-based chatbot that supports students through the complex process of applying for college and financial aid. This paper presents findings from the Optimizing Texting Technology through Engagement Research with Students (OTTERS) project, a three-phase effort to improve OtterBot’s reach and impact using a behavioral science lens and rigorous testing.The study is grounded in a central hypothesis: that improving the design of automated AI messages—to better align with students’ informational needs, address behavioral barriers, and build trust—can measurably increase engagement and drive critical outcomes like FAFSA/WASFA completion. Secondary hypotheses explore whether such effects vary by subgroup, particularly across gender, GPA, and geography.The paper draws on results from a randomized A/B test conducted during the 2023–24 academic year, in which over 10,000 students were assigned either to receive WSAC’s standard messages or a newly developed “alternative messaging” stream. These revised messages were informed by human-centered design, focus groups, and user experience surveys, and incorporated best practices in behavioral science and digital communications.Findings show that the alternative messaging significantly increased student engagement with the chatbot and produced a statistically significant increase in financial aid application completion among male students—a group disproportionately underrepresented in aid take-up. While overall application rates remained low due to broader systemic challenges (including the troubled rollout of the 2024–25 FAFSA), the experimental results suggest that well-designed AI interventions can meaningfully shift outcomes for targeted groups.The paper also highlights insights from implementation research conducted during earlier OTTERS phases. These include documentation of low baseline engagement (only 12% of College Bound seniors responded to any OtterBot message) and widespread confusion about state aid programs. Many students distrusted automated messaging and misunderstood eligibility for aid—even when guaranteed. This context informed several strategic messaging adjustments, including clearer explanations, earlier outreach, and culturally responsive framing.Finally, the paper discusses implications for policy and practice. The phased approach used in OTTERS—combining user research, AI message redesign, and experimental testing—offers a scalable model for other states seeking to improve educational equity through digital tools. It also contributes to the growing body of research on how AI-enabled outreach can advance public goals when developed through cross-sector collaboration between government, researchers, and technology partners.In line with APPAM’s 2025 theme, this project demonstrates how integrating state administrative data, behavioral research, and artificial intelligence can build evidence and tools to support equitable access to higher education.

Author

• Richard Hendra, MDRC
  Presenting Author