This paper describes an implemented method to generalize an example tutor interaction into a query to retrieve similarly related sets of events. It infers WHERE clauses to equate repeated values unlikely to match by accident. The Session Browser [1] shown in Figure 1 is an EDM tool to view data retrieved by querying a database of events logged by a tutor. It displays retrieved events in a context tree of enclosing events, with a 1-line summary of the database record for each event. Figure 1: Event context tree highlighting two events selected by the user to AutoJoin This brief example occurred in an activity to teach children to ask themselves questions about the text they read. The first highlighted event summarizes a child’s multiple-choice response to a prompt to fill in the rest of a question about the text. The second event describes the child’s spoken response to a prompt to speak the completed question aloud. Often we want a query to retrieve examples similar to a current example. Complex queries are hard to construct, so we developed AutoJoin to generate them automatically. AutoJoin generalizes the two highlighted events into a query that finds “similar” cases, in this case a multiple choice step immediately followed by a free-form response step: