Learning and Updating User Models for Subpopulations in Persuasive Argumentation Using Beta Distributions

Persuasion is an activity that involves one party (the persuader) trying to induce another party (the persuadee) to believe or do something. It is an important and multifaceted human facility both in professional life (e.g., a doctor persuading a patient to give up smoking) and everyday life (e.g., some friends persuading another to join them in seeing a film). Recently, some proposals in the field of computational models of argument have been made for probabilistic models of what the persuadee knows about, or believes. However, they cannot efficiently model uncertainty on the belief of individuals and cannot represent populations. We propose to use mixtures of beta distributions and apply them on real data gathered by linguists. We show that we can represent the belief and its uncertainty using beta mixtures and that we can predict the evolution of this belief after an argument is given. We also present examples of how to use the mixtures in practice to replace general belief update functions.