Predicting Persuasive Effectiveness for Multimodal Behavior Adaptation using Bipolar Weighted Argument Graphs.

Research states that persuasion is subjective. Moreover, people use behavioral cues all the time, very often even without noticing and are often not aware of being persuaded by non-rational cues. In order to draw attention to these effects, we want to enable virtual agents to adapt their behavior during interaction to the listener in order to increase their perceived power of persuasion. In this paper, we introduce a novel multi-modal persuasive AI system that presents arguments from an underlying logical argument structure to a user by means of a virtual agent and synthetic speech. In doing so, the agent is able to adapt its multimodal behavior to the user, based on his or her explicit feedback. To this end, the feedback is used to predict the current user's stance by considering the underlying argument structure using bi-polar weighted argument graphs to later optimize the adaptation of the multimodal presentation by means of Reinforcement Learning. We report on results of a user study with 48 participants showing the validity and practical potential of the proposed prediction model and conclude by providing limitations and implications in detail.