exploratory graph analysis
Definition
Exploratory graph analysis is a network psychometric method used to estimate the dimensional structure of psychological data by modeling variables as nodes and their statistical relationships as edges. In its dynamic form, it extends this approach to multivariate time series, revealing how variables co-vary and change across time rather than treating them as static traits. Applied to facial expression recognition scores drawn from 220 YouTube videos of global political leaders, dynamic exploratory graph analysis identified a two-dimensional network structure distinguishing positive from negative emotions, both in terms of co-occurrence and rate of change. The method also detected structural differences between leaders with varying degrees of populist rhetoric, including greater autonomy in anger expression among more populist speakers.
Sources: Tomašević & Major (2024)
Related Terms
- network analysis (1 shared article)
- emotions (1 shared article)
- populism (1 shared article)
- affective dynamics (1 shared article)
- computational social science (1 shared article)
Applications
Exploratory Graph Analysis and Emotion Dynamics
Dynamic exploratory graph analysis is directly suited to studying emotion dynamics because it models the covariation and rate of change of emotions across time as a network rather than reducing them to aggregate scores. Applied to facial expression data from political speeches, the method revealed that positive and negative emotions form distinct clusters, and that anger shows a negative correlation with most other emotions, indicating a more autonomous trajectory.
Sources: Tomašević & Major (2024)
Exploratory Graph Analysis and Populism
Differences in network structure estimated through dynamic exploratory graph analysis map onto variation in populist rhetoric among political leaders. More populist leaders show a less connected emotional network overall, with anger becoming more autonomous and happiness growing more contingent on the co-expression of other positive emotions, pointing to a strategic patterning of emotional expression.
Sources: Tomašević & Major (2024)
Exploratory Graph Analysis and Facial Expression Recognition
In the study of political communication, dynamic exploratory graph analysis was applied directly to facial expression recognition scores extracted via zero-shot machine learning from video footage. This combination allowed for the modeling of emotional co-occurrence patterns across 220 videos, offering a computational pathway for analyzing non-verbal affective content in political speeches at scale.
Sources: Tomašević & Major (2024)
