Using Virtual (3D) Models to Enhance Teaching of Rock Identification in the Geology Classroom: An Application of Education Principles Gleaned from the Cognitive Science of Category Learning

This project aims to enhance the teaching of rock identification and categorization in introductory undergraduate geoscience classes. Helping students learn about categorization, such as rock types in geology, is important in science curricula because categories are building blocks of basic thought, reasoning and inference processes.

We are conducting a multi-year classroom experiment in an introductory physical geology laboratory course that compares three conditions for teaching rock identification. In Condition 1 students are trained in a traditional way, using a single physical sample of each rock type (rocks-in-the-box). In Condition 2 students are trained in a computerized study-test procedure using four interactive 3D models of each rock type. The 116 3D models used in this training condition were created, hosted and deployed to student laptops by Rockviewer, a new software system. Condition 3 uses hybrid training with both physical samples and the 3D models. Students spend equal time training under each condition. In all conditions, following training, students are tested on their ability to classify novel physical samples of the rock types. Results obtained thus far indicate that students identify novel physical samples about equally well after training with physical samples only and after training with 3D models only. The trend is for performance to be best in the hybrid-training condition, although a larger sample size is needed to confirm statistical significance. This trend is in line with education principles gleaned from the cognitive science of category learning: Training with physical samples provides students with embodied knowledge resulting from direct manipulation of the rock types. Training with multiple virtual 3D models provides students with variability in the samples of the rock types that they experience; such variability enhances subsequent generalization to novel members of the categories. The virtual-3D models training procedure also provides students with types of retrieval practice known to enhance subsequent test performance. Our findings and technology have the potential to serve students in environments that lack physical geologic samples for learning rock identification. Supported by NSF-IUSE #1937389 and #1937361.