Ego biases in cognitive representations of social structures

Individuals' cognitive representations of their social structure are investigated in two studies utilizing a three-mode cognitive data collection paradigm due originally to Newcomb and recently developed by Krackhardt. The paradigm requires each actor in the social network to produce a digraph of the network relation. The networks studied for the present research were a group of hardware design engineers who worked together for an extended period of time at a large company, and a group of residency hall advisors who worked together at a major university housing complex. Respondents in both groups reported directed friendship ties among all actors embedded in the network, including themselves.;Each data set was analyzed using both multidimensional scaling and graph-theoretic methods. First, using descriptive correspondence analysis, data provided by each actor were scaled simultaneously into the same multidimensional Euclidean space. This approach provided a separate representation of each actor's own cognitive mapping of the structure along with a global aggregate view in the same space which facilitated comparisons among individuals' representations and their relationships to the group-level aggregate view.;The results from both studies revealed that actors have a tendency to perceive their social world in an egocentric way. Evidence is presented for ego's tendency to perceive self position as more "central" within the network compared to others' perceptions of ego. Also, ego has a tendency to partition the network into two subnets and to employ different strategies in constructing each subnet's representation. Within the innet, consisting of ego's friends, ties are perceived to be dense, mostly reciprocated and transitive. Within the outnet, consisting of all others, ties are sparser, with observed reciprocity and transitivity close to chance levels.;In addition to spatial methods, graph-theoretic centrality methods were used to uncover the biases present in individual data. Overall, the results cast doubt on the sociometric practice of constructing an aggregate social network by requiring each actor to name only their own network ties. Hopefully, the work will lead to further investigation of individual biases and knowledge differences in social perception and to more proper aggregation techniques to reveal group-level representations of the social structure.