Using Krackhardt Data Over Time to Assess Egonet Dynamics and Accuracy

Abstract

There has been some debate over the extent to which egonets support accurate inferences regarding the actual structural features of social networks. The reliance of the egonet method on ego’s accurate perception of the local network naturally motivates the examination of differences between what ego perceives and what actually exists. Three decades ago, Krackhardt pioneered the assessment of ego’s perception of alters and ties with the contribution of cognitive social structures, henceforth referred to as Krackhardt data. I claim that what has been needed is further examination of Krackhardt data over time, where possible compared with measurement of the actual network structure. This could yield progress for the field in better understanding the stability of both social network perceptions and their accuracy with respect to the actual social landscape. The work presented in this session is an example of just such an effort. My longitudinal dataset consists of the two relations of friendship and advice measured over a six week period, collected as relationships formed among a group of college juniors working together in a class that included a Prisoner’s Dilemma exercise. Notably, this exercise held the potential of significantly affecting not only the group’s actual social networks but also the participants’ perceptions of each other’s’ social networks. Augmented by Krackhardt data also collected each week during the same time period, this dataset offers some advantages over the field’s most widely analyzed dataset due to Newcomb. Analytical results presented will include longitudinal results for ego distortion, local-global inference error, and feature inference error.