TITLE: Identifying causal structure in scientific explanatory essays SPEAKER: Peter Hastings (DePaul University) ABSTRACT: Educational standards have put a renewed focus on strengthening students' abilities to construct scientific explanations and engage in scientific arguments. Because evaluating student writing is so time-intensive, people often turn to automatic essay evaluation programs, but these generally rely on shallow measures of word choice and coverage of a topic. Such methods can produce an overall, holistic evaluation of the essay quality, but do not reveal the individual connections between causes and effects that students make. Our focus is on determining the causal structure in students' explanations of scientific phenomena, so that we can provide feedback which will help them improve their explanation-writing skills. Using hand-annotated essays from thousands of high school students in the Chicago area, we have developed machine learning techniques to identify both the conceptual elements in student essays and the causal connections between them. This talk will describe the techniques that we have used and our evaluation of them.