51-2 Investigating causal analogy generation in introductory geoscience courses

Session: Geoscience Education Research: Methods, Frameworks, and Results from Emerging Scholars

Presenting Author:

Authors:

Abstract:

Ensuring students develop a deep understanding of key geoscience topics remains a challenge in introductory coursework, especially in online and asynchronous learning environments. One promising approach is the use of analogies, particularly causal analogies that relate complex underlying mechanisms and concepts in geoscience to everyday experiences. Analogy generation tasks have shown potential to enhance understanding; however, most prior work has been conducted in laboratory settings, leaving questions about their effectiveness in learning environments. 

This study leverages the Analogy Generation Task (AGT) developed in Goldwater et al. (2021), which revealed that only 16% of novices generated accurate causal analogies when prompted. We implemented the AGT in asynchronous online introductory Earth Science courses to investigate how novices compared to college students and whether certain geoscience topics yielded more analogy generation. Students in four course sections (N = 308) completed two AGT prompts focused on the causal mechanisms in magma cooling and plate subduction. Two of the course sections engaged in practice utilizing the AGT.  Comparisons with the Goldwater et al. dataset reveal no significant difference in causal analogy generation rates for the density-related AGTs between lab novices and college students (X²(1, N = 229) = 0.9616, p = 0.327). Responses from college students with practice were significantly more likely to be causal analogies overall (X²(1, N = 601) = 5.814, p = 0.016); however, this effect is not consistent across all topics, as students with practice were significantly more likely to produce a causal analogy for the magma-related AGT (X²(1, n = 303) = 7.394, p = 0.007), but showed no significant difference for the subduction prompt (X²(1, n = 298) = 0.945, p = 0.331). This may suggest that the effectiveness of analogy intervention depends on the complexity or familiarity of the underlying causal concepts.  

A challenge emerged with the prevalence of AI-generated responses, which featured similar language and vague explanations. It is difficult to determine with certainty what is and what is not artificially generated, introducing a new limitation in measuring learning outcomes with open-ended tasks, particularly in fully remote spaces.  

Future work aims to measure the impact of peer-to-peer interaction through discussion posts utilizing AGT and how student outcomes may differ when an AGT is included in a traditional homework assignment.

Geological Society of America Abstracts with Program. Vol. 57, No. 6, 2025

© Copyright 2025 The Geological Society of America (GSA), all rights reserved.