Temporal Social Network Analysis of Discourse

In this paper, we explore a number of techniques for two-dimensional visualisation of temporal social network data, with the goal of providing feedback on the group dynamics of government planning workshops. These techniques include the use of position and colour for displaying temporal information, as in colour-coded bar charts or sequence diagrams. While it is possible to display temporal information solely using colour, experiments with an Internet discussion group showed that the most useful techniques are plots with an explicit horizontal time axis, and a vertical indication either of the amount of communication by actors (weighted degree), or of the importance of actors in the patterns of interaction (centrality). We have applied these visualisation techniques to interaction data from a workshop case study (Fleming, 2008). Doing so identified scaled centrality plots, such as that in Figure 1, as particularly useful for providing insights into the changing patterns of interaction between workshop participants. These plots display actor centrality calculated over a sliding window, scaled by the average centrality in that window, and are useful for providing feedback to workshop facilitators on the overall “flow” of their workshop over time. Figure 1. Scaled centrality plot for our workshop case study. The nine sub-phases of the workshop (discussed in Section 4) stand out clearly, making such plots helpful in visualising workshop dynamics. In particular, there is a very brief initial dominance by the facilitator (setting the scene), followed by a period dominated by participant P5, then by renewed facilitator control, then by a discussion dominated by P3 and P13, then by five other sub-phases. The explicit time axis assists in communicating the information in this diagram to workshop facilitators.