Search for: All records

Understanding how novices acquire and hone visual search skills is crucial for developing and optimizing training methods across domains. Network analysis methods can be used to analyze graph representations of visual expertise. This study investigates the relationship between eye-gaze movements and learning outcomes among undergraduate dentistry students who were diagnosing dental radiographs over multiple semesters. We use network analysis techniques to model eye-gaze scanpaths as directed graphs and examine changes in network metrics over time. Using time series clustering on each metric, we identify distinct patterns of visual search strategies and explore their association with students’ diagnostic performance. Our findings suggest that the network metric of transition entropy is negatively correlated with performance scores, while the number of nodes and edges as well as average PageRank are positively correlated with performance scores. Changes in network metrics for individual students over time suggest a developmental shift from intermediate to expert-level processing. These insights contribute to understanding expertise acquisition in visual tasks and can inform the design of AI-assisted learning interventions. 

This study aims to systematically evaluate the use of social network analysis (SNA) metrics to measure eye-tracking behavior to assess and predict student learning performance. We integrated 11 network metrics from published research and tested them on six eye-tracking datasets. Our preliminary results indicate that no consistent predictor variable can effectively predict student performance across different datasets. The number of nodes, edges, reciprocity, and entropy measures contribute differently to predicting students’ performance. This work deepens our understanding of how different SNA metrics relate to eye-tracking data and advances the methodological framework to predict learning outcomes. 

When students reflect on their learning from a textbook via think aloud, network representations can be used to capture their concepts and relations. What can we learn from these network representations about students’ learning processes, knowledge acquisition, and learning outcomes? This study brings methods from entity and relation extraction using classic and LLM-based methods to the application domain of educational psychology. We built a ground-truth baseline of relational data that represent relevant (to educational science), textbook-based information as a semantic network. We identified SPN4RE and LUKE as the most accurate method to extracting semantic networks capturing the same types of information from transcriptions of verbal student data. Correlating the students’ semantic networks with learning outcomes showed that students’ verbalizations varied in structure, reflecting different learning processes. Denser and more interconnected semantic networks indicated more elaborated knowledge acquisition. Structural features such as the number of edges and surface overlap with textbook networks significantly correlated with students’ posttest performance. 

Student during-learning data such as think-alouds or writing are often coded for use of strategies or moves, but less often for what knowledge the student is using. However, analyzing the content of such products could yield much valuable information. A promising technique for analyzing the content of student products is semantic network analysis, more widely used in political science, communication, information science, and some other social science disciplines. We reviewed the small literature on semantic network analysis (SemNA) of individuals with relevant outcomes to identify which network analysis metrics might be suitable. The Knowledge Integration (KI) framework from science education is discussed as focusing on amount and structure of student knowledge, and therefore especially relevant for testing with SemNA metrics. We then re-analyze three published think-aloud data sets from undergraduate students learning introductory biology with the metrics found in the literature review. Significant relations with posttest comprehension score are found for number of nodes and edges; degree and betweenness centrality; diameter, and mean distance. Inconsistent results possibly due to text-specific features were found for number of clusters, LCC, and density, and null results were found for PageRank centrality and centralization degree. Basic principles from the KI framework are supported—amount of information (nodes), connections (edges, average degree), key ideas (degree and betweenness centrality) and length of causal chains (mean distance and diameter) are related to posttest comprehension, but not density or LCC. Possible explanations for slight variations across data sets are discussed, and alternative theories and metrics are offered.