B0601
Title: Agent-based null models for examining experimental social interaction networks
Authors: Kevin Burke - University of Limerick (Ireland) [presenting]
Abstract: The analysis of temporal data is considered, arising from online interactive social experiments. The analysis of such data is complicated by the fact that observations are interrelated since participants are exposed to the same social experience. Therefore, an approach is proposed that generates a null distribution for fitted linear regression coefficients based on an underlying agent-based model; the particular interest is in the null model of participants interacting at random. In addition to this, network visualisations are provided that characterise a given experiment and identify individuals whose behaviour is atypical. The experimental data that is considered has been collected using a virtual interaction application (VIAPPL), wherein participants interact with each other over a series of rounds. In this context, it is found that: participants prefer to interact with participants assigned to the same group as them; participants reciprocate with each other; and these behaviours strengthen over the course of the experiment. Although the proposed approach has been developed with VIAPPL in mind, it is sufficiently generic that it could be used with other forms of social interaction data.
