Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Using network science to provide insights into people’s understanding of activity centrality

Poster Session E - Monday, March 9, 2026, 2:30 – 4:30 pm PDT, Fairview/Kitsilano Ballroom

Mackenzie Bain¹ (), Martha Valmana Crocker¹, Beatrice Valmana Crocker¹, Kara E. Hannah¹, Kevin S. Brown², Ken McRae¹; ¹University of Western Ontario, ²Oregon State University

An important issue in event cognition concerns how activities such as “order a drink” come to mind when people think about events like eating at a restaurant. Human empirical research has demonstrated that an activity’s centrality (importance) influences how people think about activities and events. However, no implemented computational measures have been used to predict or provide insights into people’s understanding of activity centrality. We used network science to construct 80 temporally structured event networks from human participant ordering and production data, and derived quantitative predictions from those networks. We used five network centrality measures computed from the ordering and production networks: CheiRank, PageRank, 2D Rank, Betweenness, and Closeness. Participants ranked activity centrality on one of five measures: centrality, standardness, surprise, prominence, and goodness of fit. Linear mixed-effect regression analyses showed that CheiRank, which assigns importance to activities that have influential outgoing links leading to other (later) influential activities, was the strongest predictor of participants’ rankings for all five centrality measures. Closeness, which assigns importance to activities that have short paths to them from an event’s earlier activities, was the second strongest predictor for all five measures. The remaining network centrality measures did not predict participants’ rankings. Together, CheiRank and Closeness accounted for a large proportion of variance in participants’ rankings with R2 ranging from .23 to .34 across the five measures. These results show that people’s perception of centrality can be predicted by a model that captures the rich and variable nature of the temporal structure of events.

Topic Area: LONG-TERM MEMORY: Semantic