Infants of slow-developing, altricial species must learn crucial information and skills from their caregivers and other adults. How do infants identify social partners and learn from them? The challenges inherent in learning from social interactions include: identifying social partners, learning the structure of others’ behavior, coupling one’s perceptual and motor actions to those of the social partner, and recognizing and attributing agency and intentionality. Our workshop will illustrate mechanisms at multiple levels of organization that are involved in meeting these challenges. The goal of the workshop is to connect research in developmental psychology, animal behavior, developmental robotics, and computational modeling to promote a more integrative understanding of learning in social contexts.
Findings discussed in this workshop have the potential to lead to theoretical breakthroughs in understanding human (and animal) development and learning, and also have applied utilities in human-robot interactions. Several of the workshop speakers are at the forefront of a new trend in cognitive science, studying micro-level behaviors (e.g. eye gaze and body orientation) in their investigations of learning and development of complex natural organisms in natural environments. With advances in computer vision, speech processing and machine learning, we now have capabilities to process visual, audio and other sensory data collected from naturalistic behaviors from relatively naturalistic environments. In this way, we can measure and analyze the physical and social regularities of the real world as experienced by a young learner. Such an understanding of the learning environment provides unique opportunities to study underlying mechanisms of cognition and learning. In addition, those micro-level behavioral patterns capture the real-time dance of dynamic adaptive behaviors between two social partners. Understanding the interactions between brain, body and environment, and across multiple social participants, allows us to test and falsify formal theories about specific emerging cognitive functions, and to measure and analyze a rich history of interaction in real time and space within highly diverse environments.
Moreover, artificial systems (e.g. physical robots and virtual humans) can interact with people in everyday contexts, which not only provides a way to enrich experimental methods but also has applied utilities. For example, embodied models (e.g. baby robots) may learn from interacting with adult users (e.g. caregivers). In addition, teaching robots can interact with young children (in particular, atypical development groups) to facilitate their social skills and learning. A deep understanding of behavioral patterns and underlying cognitive principles in social contexts has special advantages for transfer to human-robot interactions as those patterns can be directly incorporated into robot sensory and control systems to lead to more human-like and/or human-friendly behaviors in human-robot interactions.
Target audience. The workshop will benefit all disciplines within the ICDL-Epirob community, including developmental psychology, cognitive neuroscience, computational modeling, and developmental robotics. These fields represent crucial areas of investigation into the challenges of learning in social interactions.