Towards Mental Model-driven Conversations

In recent years conversation has become a key channel for human-computer interaction. Dialogue personalization could result in an important aspect, making sense of users' features when engaged in a conversation with a machine. A feature that has been properly taken into account is the user's mental model, a crucial aspect since it determines users' expectations and the way they interact with a chatbot. In this position paper, we propose a theoretical framework that combines existing meta-mental models (behaviour-based and lexical-based ) in a computational model that can be used to automatically detect the users' mental model from the dialogues with a chatbot by exploiting Linguistic theory and Machine Learning techniques.