AI for Identifying Dismissive and Acceptive Acts in a Conversation Dataset and model code

Human interactions involve dialogue acts that can be responded to by acceptive or dismissive acts. For example, a person sharing a painful past event uses the dialogue act of disclosure. The disclosure can be responded to empathically or by dismissing the pain. Here, we address the challenge of automatically identifying dismissive and acceptive dialogue acts in a conversation. We used massive AI-generated datasets of utterances and dialogues expressing acceptive/dismissive behavior to address the challenge. Next, we trained and tested several machine-learning models that performed highly in classifying utterances as acceptive or dismissive. The basic approach described in this paper can empower the development of automatic interactive systems in contexts ranging from artificial therapists to assistant robots for the elderly.