June 15, 2017

This session introduces a study of search algorithms in legal databases (Casetext, Fastcase, Google Scholar, Lexis Advance,Ravel,and Westlaw). Their algorithms are created by humans, who have made multiple choices about how the search will be carried out and what processes and documents will be used to enhance search results. Legal researchers don’t know a lot about what those choices were, but the choices have a dramatic effect on the search results each database returns. Every algorithm works in a unique matrix of assumptions, biases, and enhancements, that might be called an algorithmic world view, and those world views vary widely. The study used identical searches in identical jurisdictional databases in each of the six legal databases, and the variations in results are startling. Every legal database returns unique and relevant results, and the overlap in the results is much less than might be expected. One major takeaway: we need keep asking for more algorithmic accountability from database providers.