Simulation of User Participation and Interaction in Online Discussion Groups

The main objective of this paper is to study through agent-based simulation the process in which an online discussion group as a continually changing virtual community behaves over time. In a previous study two different international online discussion groups were analysed by means of grooming analysis. Grooming analysis is a method that use the concept of groom as a social reward in the form of attention that one group member gives to another. By means of analysis of how attention was distributed it was possible to estimate the social status ranking in the groups. This estimation showed to correlate well with that made by a group of independent evaluators in a validation (correlation = 0.84) It was also found that receiving grooms was an important factor that determined if a group member would stay in the group or not, and also how many posts would be made. In this second study an agent-based simulation model was constructed in Mathematica that could replicate the time dependent behaviour of the two groups. Experimentation with model parameters showed that random selection of group members for posting did not produce the observed behaviour. A selection method based on a simulated "wit-value" could not replicate the observed behaviour. However, selection methods based on counting the number of grooms received gave a good fit to observed data. A comparison was also made between the predictions made by Wilkinson.