Applying Computational Methods to the Study of Commonsense Science Knowledge

sherin 

Wednesdays@NICO Seminar, Noon, April 29 2009, Chambers Hall, Lower Level

Prof. Bruce Sherin, Northwestern University

Abstract 

Over the last three decades, researchers in science education have devoted substantial effort to the study of commonsense science knowledge, the informally-gained knowledge of the natural world that students possess prior to formal instruction in a scientific discipline. Research in this tradition has, to date, generally relied on hand-coding of interview data. In this talk, I will discuss our attempts to automate the coding of data; beginning with data in the form of raw interview transcripts, I will show how it is possible both to induce coding categories and code the transcripts, all without supervision from a human coder. Given the tacit assumptions that are built into the field’s theories and methods, this result is quite surprising. In this work, we make use of a data corpus consisting of clinical interviews in which middle school students were asked to explain the seasons. The computational techniques I will describe are principally based on a combination of Latent Semantic Analysis and Cluster Analysis, and build on initial work, using the same corpus, by Gregory Dam and Stefan Kaufmann (Dam & Kaufmann, 2008).