Taking a stance: experimenting with deliberation in dialogue

Abstract

Abstract How do people manage disagreements in conversation? Previous studies of dialogue have shown that the interactional consequences of disagreement are not straightforward. Although often interpreted as face-threatening when performed in an unmitigated manner, disagreement can also encourage novel contributions. This thesis explores how systematically altering the presentation of someone’s stance influences the deliberative potential of a dialogue. A corpus analysis of ordinary conversations shows that exposed disagreement occurs rarely, but that speakers can signal a potentially adversarial position in a variety of other ways. One of the most interesting among these is the way people mark their rights to speak about something. Resources such as reported speech and prefacing incongruent content with discourse markers (e.g. ‘well’) can be important to the management of interpersonal factors. The idea that disagreement is problematic but also useful for deliberation is examined. Using a method that allows fine-grained manipulations of text based dialogues in real-time, agreement and disagreement fragments are inserted into a discussion dialogue. The findings show that inserting exposed disagreement violates the conventions of polite dialogue leading participants to put more effort into the production of their replies, and does not improve levels of deliberation. This raises the question of whether manipulating apparent degrees of speaker commitment might be more important for influencing the quality of deliberation. An experiment was devised which presented oppositional content with differing degrees of ‘knowingness’. The findings indicate that marking stance as knowing leads to less guarded exchanges, but does not increase deliberation. Conversely, framing statements as less knowing increases the likelihood that participants consider more alternative viewpoints, thus increasing the deliberative quality of a dialogue. Potential applications include training guidelines for professionals developing tools to support considered debate. Implications for computational argumentation studies include the importance of interpersonal dynamics and stance construction for formulating polite arguments.