LAUSR

.edu When I started my job as research director of the Massachusetts Department of Elementary and Secondary Education twelve years ago, I thought my job was to figure out what worked. My agency was just beginning to have access to new and exciting longitudinal data on students and educators. I envisioned that I’d use those data along with strong designs for causal inference to determine which programs and policies were working and which were not. Once we knew those answers, I figured, we would get better policy that would improve outcomes for Massachusetts’ students. But in my twelve years in this job, I’ve learned that the process of improving policy1 through research is much subtler and more complex than I had initially imagined. Research influences policy more often than much of the academic community thinks, and more frequently every day as we learn how to do this work better. But its influence is less linear than researchers expect, and it is driven as much by relationships and organizational capacity as by the actual information studies produce. Research use operates through conversations, not code; structures in organizations, not standard errors; relationships, not randomized controlled trials. I worry that the growing national efforts to connect research and policy too frequently start from the same “find what works” frame of mind that I did twelve years ago. The “find what works” approach misunderstands the problem of research use as one of lack of information—either lack of information about the impact of a policy or lack of awareness by the policy maker about the available information—and a need for “translation” across sectors (Penuel et al. 2015). This belies the research literature about how research actually plays into the policy decision process. If