The Social AI Group and the Social AI Centre for Doctoral Training (CDT) are organising the Third Workshop on Artificial Social Intelligence (Social AI) on 26th September 2025 at the Researcher Development Rooms (226A&B), Advanced Research Centre, University of Glasgow, UK. 

Social AI involves developing an AI domain aimed at endowing artificial agents with social intelligence, the ability to deal appropriately with users’ attitudes, intentions, feelings, personality and expectations. This full day workshop (9am to 4pm) will host a series of invited talks by renowned experts in Social AI. Our goal is to bring together academic experts, students and industry professionals to encourage dialogs around the progress, challenges and opportunities in Social AI as AI continues to permeate all aspects of our social presence. 

Keynote Speakers include:

Prof. Albert Salah, Utrecht University

"Designing computational tools for behavioral and clinical science"

Automatic analysis of human affective and social signals brought computer science into closer alignment with social sciences, enabling new collaborations between computer scientists and behavioral researchers. In this talk, I highlight the key research directions in this burgeoning interdisciplinary field, and provide an overview of its major opportunities and challenges. Computer science and psychology have different methodological assumptions and approaches. Drawing on examples from our recent research - such as automatic analysis of interactive play therapy sessions with children, and diagnosis of bipolar disorder from multimodal cues - as well as relying on examples from the growing literature, I explore the potential of human-AI collaboration, where AI systems do not replace, but support monitoring and human decision making in behavioral and clinical sciences. In particular, the role of face, body, gesture, speech and multimodal analysis  are discussed, as well the role of explainability and interpretability, which are important aspects for trustworthy computer systems in this domain. 

Arabella Jane Sinclair, University of Aberdeen 

"Repetition and adaptation in humans and language models"

From children echoing caregivers to learn how to form utterances, to second-language learners mirroring teachers to gain fluency, to collaborators navigating knowledge asymmetries to ground goal-oriented dialogue, repetition shapes how we communicate and coordinate. This talk examines how we repeat and adapt to one another in dialogue, exploring the multiple functions of repetition in conversational interaction, including easing processing demands, facilitating grounding, providing feedback, and signalling social alignment. I will show that repetition and adaptation in human-human dialogue occurs across different levels of communication—lexical, structural, and gestural; that repetition is local in scope; varies with speaker relationships and communicative abilities; and can facilitate communicative success. 

I will then turn to repetition and adaptation in Language Models, the backbone to many human-facing conversational technologies. When generating next utterances within a dialogue context, LMs mirror some of the repetition behaviour associated with efficient collaborative dialogue in humans, including local repetition of lexical and syntactic forms. Moreover, in a behavioural task setting similar to priming studies in psychology, LMs’ expectations about upcoming structural material are modulated by similar contextual cues as in humans.

In the final part of this talk, I will turn to human-machine interaction, focusing on how LM-based conversational systems could adapt to speakers from different social groups. I will also consider how, in the absence of system adaptation, human interlocutors modify their own language to accommodate the system—sometimes to a greater extent than when interacting with other humans. Overall, this research highlights the importance of the subtle dynamics of repetition and adaptation for both human communication and the development of social AI.

Prof. Alessandro Vinciarelli, University of Glasgow

Social AI at UofG: History, Impact and Collaboration Opportunities

The goal of this talk is to introduce Social AI, the field aimed at making machines socially intelligent, i.e., capable to automatically make sense of human-human and human-machine social interactions in the same way as people do. The presentation will cover the main scientific and technological questions underlying the field and, in addition, while also showcasing the main Social AI activities conducted at the University of Glasgow since 2010, when the field was still in its pioneering stage. The focus will be on the development of a thriving Social AI research community in the University -- a journey that culminated in the establishment of a UKRI Centre for Doctoral Training (CDT) in Socially Intelligent Artificial Agents, the only CDT in AI awarded to the University of Glasgow to date. Furthermore, special emphasis will be placed on the rich opportunities for interdisciplinary collaboration that Social AI offers, especially in terms of interaction between AI experts, psychologists, psychiatrists, sociologists, neuroscientists and all other researchers interested in understanding human behaviour and its underlying phenomena.