A cognitive approach to modeling bad news conversations

Abstract

One of the fundamental aspects of human culture is our capacity to use complex natural language to communicate with each other. The complexity of language is caused by the huge diversity in the use and meaning of words, gestures and facial expressions. The research that is presented in this thesis has several focus points. First, it focuses on how the function and meaning of verbal and non-verbal communicative behaviors in medical bad news situations can be determined. We show that these two aspects can, for a large part, be determined by analyzing the properties and characteristics of conversational behaviors. Analysis of the properties and characteristics was done based on the study of various linguistic and psychological theories and models such as Speech Act Theory and Dialogue Act Theory. Secondly, it focuses on determining how perceived conversational behaviors are cognitively processed by a listener and how the listener generates an appropriate response behavior. To this end, several approaches that describe features of the internal states of people (including beliefs, desires and intentions, but also emotions and social dispositions) are presented. In additions, we argue that the listener’s assumptions about the internal state of the speaker are essential in the processing mechanisms. The third focus of the research is to determine how the cognitive processes and the associated internal state features might be represented. To that end, a cognitive dialogue model was constructed where effort was made to keep the representation of the processes and features as natural and human-like as possible. The idea behind this is that, based on the cognitive dialogue model, a dialogue system can be constructed that performs human-like conversational behaviors while at the same time have underlying processes that can be intuitively be understood by the user.