About this Research Topic
The coming years might give rise to the diverse needs of socially intelligent conversational agent systems that are required to cooperate with humans in virtual and physical settings, such as healthcare, life-long education, group work, entertainment, and so forth. Therefore, how should socially intelligent conversational agents be designed to coexist in harmony with people and expand the capabilities of society as a whole? They might need abilities to understand human users' intentions and beliefs and to help them achieve their goals. Such agents might also have to adapt to individual and socio-cultural norms and protocols, and to establish long-term trust by being accountable, and behaving predictably and consistently.
Because of the nature of conversations, socially intelligent conversational agents require holistic system integration with modules of understanding, reasoning, learning, decision making, generation, and any other functions that enable social interactions. Although dialogue services for specific tasks, such as smart speakers and personal assistants, widely benefited from the emergence of deep learning and recent advances in the field of natural language processing, fewer efforts have been carried out to design, build, and integrate complex socially intelligent conversational agent components altogether. Thus, what kind of design principles and computational models would be needed to achieve the unification of such complex functions? Neuro-symbolic and cognitive architectures could be conciliatory and promising approaches towards the construction of conversational agents that exhibit social behavior guided by a theory of mind, be capable of genuinely performing NLU (not just NLP), build shared mental models to support human-machine teamwork, and perform social context understanding.
To make a solid step toward the realization of socially intelligent conversational agents, it is crucial to identify proper tasks/domains with concrete evaluation functions. The success of Google DeepMind's AlphaGo has been attributed to the fact that they chose a domain with clear and unambiguous rules (the game of Go). What about the requirements for the task design for the growth of socially intelligent conversational agents? Once a task is determined, how to establish a data collection ecosystem in which data can be sustainably gathered, and hypotheses can be iteratively tested at scale?
In this Research Topic, we welcome explorative and innovative submissions led by scientific intuition in the field of socially intelligent conversational agents and their applications. Areas to be covered in this Research Topic may include, but are not limited to:
● Computational models of social conversations
● Integration between task and social conversational scenarios
● Neuro-symbolic hybrid cognitive architectures for social conversations
● Non-verbal behavior generation of socially conversational agents/robots
● Situated conversational agents
● Explainable conversational agents
● Human-in-the-loop machine learning for data collection and annotation
● Applications of socially intelligent conversational agents
● Surveys and open problems of socially intelligent conversational agents
● Evaluation methods for social interactions
● Embodied conversational agents
Because of the nature of conversations, socially intelligent conversational agents require holistic system integration with modules of understanding, reasoning, learning, decision making, generation, and any other functions that enable social interactions. Although dialogue services for specific tasks, such as smart speakers and personal assistants, widely benefited from the emergence of deep learning and recent advances in the field of natural language processing, fewer efforts have been carried out to design, build, and integrate complex socially intelligent conversational agent components altogether. Thus, what kind of design principles and computational models would be needed to achieve the unification of such complex functions? Neuro-symbolic and cognitive architectures could be conciliatory and promising approaches towards the construction of conversational agents that exhibit social behavior guided by a theory of mind, be capable of genuinely performing NLU (not just NLP), build shared mental models to support human-machine teamwork, and perform social context understanding.
To make a solid step toward the realization of socially intelligent conversational agents, it is crucial to identify proper tasks/domains with concrete evaluation functions. The success of Google DeepMind's AlphaGo has been attributed to the fact that they chose a domain with clear and unambiguous rules (the game of Go). What about the requirements for the task design for the growth of socially intelligent conversational agents? Once a task is determined, how to establish a data collection ecosystem in which data can be sustainably gathered, and hypotheses can be iteratively tested at scale?
In this Research Topic, we welcome explorative and innovative submissions led by scientific intuition in the field of socially intelligent conversational agents and their applications. Areas to be covered in this Research Topic may include, but are not limited to:
● Computational models of social conversations
● Integration between task and social conversational scenarios
● Neuro-symbolic hybrid cognitive architectures for social conversations
● Non-verbal behavior generation of socially conversational agents/robots
● Situated conversational agents
● Explainable conversational agents
● Human-in-the-loop machine learning for data collection and annotation
● Applications of socially intelligent conversational agents
● Surveys and open problems of socially intelligent conversational agents
● Evaluation methods for social interactions
● Embodied conversational agents
Keywords: Socially Intelligent Conversational Agents, Long-term Trust and Rapport, Human-AI Co-Evolution, Cognitive Architectures for Conversational Modelling, Social Robots
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
