Scale-Free Superiority, Egocentric Bias and Network Centrality Heuristics in Social Graph Learning

31 Pages Posted: 22 Aug 2012  

Jason J. Jones

State University of New York (SUNY), Stony Brook; University of California, San Diego (UCSD)

Date Written: August 21, 2012

Abstract

This study examined factors predicted to affect how well the connections between people would be learned in a social graph learning task. Participants learned “who is friends with whom” in three social graphs of varying internal structure. The description of the task was varied between subjects. The structure of the graph predicted how quickly knowledge of the graph was acquired. Scale-free graphs were acquired more quickly than random graphs and “caveman” graphs. The surface description of the task had little or no effect on the speed of acquisition. Framing the task such that it seemed more personally relevant did not augment performance. Neither did making the task (hypothetically) survival-relevant to the learner. Nevertheless, within a graph, participants displayed egocentric bias in their attention and learning – they learned best about relationships which involved a node labeled “You.” Additionally, participants employed network centrality heuristics. They learned best about relationships involving nodes of extreme network centrality – those nodes that were either extremely well-connected or poorly-connected to the rest of the graph.

Keywords: learning, memory, social networks

Suggested Citation

Jones, Jason J., Scale-Free Superiority, Egocentric Bias and Network Centrality Heuristics in Social Graph Learning (August 21, 2012). Available at SSRN: https://ssrn.com/abstract=2133622 or http://dx.doi.org/10.2139/ssrn.2133622

Jason Jeffrey Jones (Contact Author)

State University of New York (SUNY), Stony Brook ( email )

Stony Brook, NY 11794
United States

University of California, San Diego (UCSD) ( email )

9500 Gilman Drive
Mail Code 0502
La Jolla, CA 92093-0112
United States

HOME PAGE: http://psy2.ucsd.edu/~jasonjones/

Paper statistics

Downloads
30
Abstract Views
358