A General Temperature Metric Framework for Conversational Interactivity

In order to investigate the potential impact of interactivity on the perceived quality of future telecommuncation services, it is necessary to define a simple but efficient metric for this conversation parameter. To this end, recently it has been proposed to use an obvious analogy to statistical thermodynamics and define interactivity as system temperature of a Markov chain describing the conversation. The present paper refines and extends this approach towards a general framework for the resulting temperature metric. We describe several state models for conversations providing for interactivity-specific requirements. Based on that, an additional parameter is assigned to each state in order to describe its specific impact on the interactivity metric properly. We derive an explicit solution of the resulting metric, before we finally discuss how to deal with the case of more than two participants by introducing a new entropy model for conversational temperature.