ChaLearn

Workshop description

Summary

Human interaction has been a central topic in psychology and social sciences, aiming at explaining the complex underlying mechanisms of communication with respect to cognitive, affective and behavioral perspectives. From a computational point of view, research in dyadic and small group interactions enables the development of automatic approaches for detection, understanding, modeling and synthesis of individual and interpersonal social signals and dynamics. Many human-centered applications for good (e.g., early diagnosis and intervention, augmented telepresence and personalized agents) depend on devising solutions for such tasks.

Verbal and nonverbal communication channels are used in dyadic and small group interactions to convey our goals and intentions while building a common ground. During interactions, people influence each other based on the cues they perceive. However, the way we perceive, interpret, react and adapt to them depends on a myriad of factors. Such factors, which we refer to as context, may include, but are not limited to: our personal characteristics, either stable (personality, cultural background and other socio demographic information) or transient (mood, physiological or biological factors); the relationship and shared history between individuals; the characteristics of the situation and task at hand; societal norms; and environmental factors. To analyze individual behaviors during a conversation, the joint modeling of participants is required due to the existing dyadic or group interdependencies. While these aspects are usually contemplated in non-computational dyadic research [1], context-and interlocutor-aware computational approaches are still scarce, largely due to the lack of datasets providing contextual metadata in different situations and populations.

To advance and motivate the research on the field, this workshop has an associated challenge using a large scale, multimodal and multiview dataset [2] recently collected by our group, that provides many challenges related to the analysis and understanding of human behaviour in dyadic interactions. As detailed on the challenge webpage, the challenge will address two different problems (divided in 2 tracks), where context and both interlocutors’ information can be exploited due to the nature of our dataset. In summary, the challenge will cover:

Automatic self-reported personality recognition (from one view per interlocutor), and
Behavior forecasting (facial landmarks, hand, and upper body pose prediction from one view per interlocutor).

All these tasks are multimodal, that is, audio-visual information and transcriptions will be given.

[1] David A Kenny. “Models of non-independence in dyadic research,” Journal of Social and Personal Relationships, 1996.
[2] Palmero et al. “Context-Aware Personality Inference in Dyadic Scenarios: Introducing the UDIVA Dataset,” Winter Conference on Applications of Computer Vision (WACV) Workshops, 2021.

Topics and Motivation

In line with these, we would like to bring together researchers in the field and from related disciplines (e.g. affective computing, social signal processing and human-computer interaction) to discuss the advances and new challenges on the topic of dyadic and small group interactions. We want to put a spotlight on the strengths and limitations of the existing approaches, and define the future directions of the field. In this context, we accept papers addressing the issues related to, but not limited to, these topics:

Detection, understanding, modeling and synthesis of individual and interpersonal social signals and dynamics;
Verbal / nonverbal communication analysis in dyadic and small groups;
Contextual analysis in dyadic and small groups;
Datasets, annotation protocols and bias discovering/mitigation methods in dyadic and small groups;
Interpretability / Explainability in dyadic and small groups;

ICCV and PMLR proceedings

Accepted peer-reviewed papers will be published in ICCV or PMLR proceedings, according to the submission choice, detailed in "submission instructions" (here).