GSA Annual Meeting in Indianapolis, Indiana, USA - 2018

Paper No. 83-2
Presentation Time: 8:20 AM


NOSOFSKY, Robert, Psychological and Brain Sciences, Indiana University, 1101 E. Tenth Street, Bloomington, IN 47405, SANDERS, Craig, Department of Psychology, Vanderbilt University, Nashville, TN 37203, DOUGLAS, Bruce J., Department of Earth and Atmospheric Sciences, Indiana Univ, 1001 E. 10th St, Bloomington, IN 47405 and MCDANIEL, Mark, Psychological and Brain Sciences, Washington University in St. Louis, One Brookings Drive, St. Louis, MO 63130

A recent theme in advancing STEM education has been to develop collaborations between cognitive-science researchers and discipline-based education research. During the past few years, our research team has pursued that theme in the domain of geoscience education. The cognitive-science researchers on the team are experts in the development and testing of formal models of human category learning; the geology experts have a long history of research and practice in geoscience education. The specific project that we have pursued involves the use of formal models of human category learning to help guide the search for effective methods of teaching rock identification and categorization. The basic idea is to simulate alternative teaching methods using the formal models themselves, and to focus empirical tests on those teaching methods that the models suggest will be most successful.

We illustrate the application of the models in a variety of highly controlled laboratory experiments in which novice students are trained to identify different sets of igneous, metamorphic and sedimentary rocks. The models are used to provide detailed quantitative accounts of the students’ learning trajectories and their ability to generalize correctly to novel samples from the trained rock types. As a prerequisite for applying the models in this manner, an initial step is to derive a high-dimensional “feature space” in which the numerous rock samples are embedded. We describe a variety of complementary methods for achieving this initial goal, including similarity-scaling studies, direct dimension-rating studies, and application of modern “deep-learning” technologies.

Finally, we describe a variety of recommendations that the models make for enhancing the teaching of rock identification; and we describe the preliminary support that we have gathered for these recommendations in our laboratory experiments.