Socio-Inspired Neural Networks for Conversational Systems – SINNet

A single lack of social tact from a conversational system (chatbot, vocal assistant, social robot) can provoke a decrease in the user’s trust and engagement towards the system. This lack of social intelligence hampers a large audience’s willingness to deem conversational systems acceptable. In order to understand the user’s state, the current research community of Affective/Social Computing has relied on research on both Artificial Intelligence and social sciences. However, in recent years, the trend has shifted to a monopoly of deep learning methods which are quite powerful yet opaque and greedy for annotated data and less suitable for the integration of social science knowledge. SINNet proposes a paradigm shift for rendering conversational systems and social robotics a more acceptable and trustworthy technology even when using deep learning approaches. It will focus on the verbal component of the interaction, will target the agent-user social relationship, and model the behaviors indexing the state of the social relationship between agent and user, going thus beyond the analysis of the user’s positive and negative sentiments. It implies developing easy-to-adapt and easy-to-explain neural models able to analyze the user’s behavior contributing to user-agent co-construction processes such as the ones characterizing the rapport with the agent, or the trust and affiliation in the agent, as well as to generate the agent’s answer fostering the user-agent social relationship. This SINNet project will establish interdisciplinarity as a core challenge by providing a shared formalism between complex (e.g., psychological or socio-linguistic) theories of social interactions and the underlying formalism in deep learning and language models.