The ultimate goal of Cognitive Developmental Robotics is to quantitatively understand and model the mechanisms of human cognition , implement these models in embodied robots, and refine the models and deepen our understanding of cognition through robot operation. Understanding these models allows for the creation of friendly and appropriate robotic intelligence that aligns with human cognition, providing significant insights for the future realization of AI robots that coexist and collaborate with humans.

Cognitive developmental robotics community have already discussed the human cognition model and its application to social robots, such as insights into the mechanisms of human attention, vocabulary acquisition based on joint attention, acquisition of context and motion concepts from action imitation, and the acquisition of a body schema based on tactile information and differentiation of self from others. We confirmed the bottom-up experiences through embodied interaction is a significant factor to understand the mechanisms for the development of intelligence that enables social communication. A typical societal contribution has been in understanding the mechanisms of Autism symptoms, where validating autism models with robots has led to understanding the causes of symptoms and the development of robots capable of interacting appropriately with autistic children.

However, a remaining challenge has been a tendency for experiments to focus predominantly on one-on-one communications, such as between children and caregivers, due to the physical limitations of using embodied robots. Understanding the mechanisms of intelligence formation based on social interactions involving more than three parties, and the development of AI to support social interactions, are important themes, yet physical constraints have tended to hinder progress.

Meanwhile, the development of modeling techniques for intelligence based on large datasets, including recent advancements in LLMs and generative AI, can complement the vast data on social group interactions that cognitive robotics has struggled with. However, the drawback of LLMs, generative AI, and foundation models is their lack of embodiment; even if multimodal information is available, these advanced technologies do not cover the function of actively performing physical actions to gather information. More in general, most generative AI approaches are based on vast amounts of data, such that open-world reasoning becomes almost in-distribution. This is feasible in contexts such as language, images or videos, but becomes particularly challenging in real-world settings. Furthermore, all the data are passively processed by the model. This represents a striking difference with respect to human cognition, which depends on an active developmental process, where much less data are actively collected, guided by the system's (i.e., the child’s) internal motivation, and leading to the discovery of causal relations. This leads in humans to the powerful capabilities of generalization and prospection that current technologies have not yet achieved. How to integrate cognitive robotics with LLMs, generative AI, and foundation models, to better understand human cognition and how to develop new research methodologies are currently crucial discussion topics.

Additionally, technologies that recreate social interactions based on physical bodies in virtual spaces like the Metaverse are becoming widespread. This technology also has the potential to overcome the constraints of traditional embodied robots, and by participating in the Metaverse, both humans and robots can potentially elucidate mechanisms for the development of intelligence based on social interactions, considering both sociality and physicality.

Thus, establishing new research methodologies for cognitive robotics, leveraging also on the new advances in technology is one of the goals of this Shonan meeting. Another goal is to create research themes that could not be studied or discussed before, based on these new research methodologies.

This workshop aims to spark an inclusive, interdisciplinary discussion on the future relationship between humans, robots, and intelligent systems. While the focus remains anchored in the traditional themes of Cognitive and Developmental Robotics, we also invite contributions from cutting-edge A.I. research, philosophy, psychology, neurophysiology, affective computing, and other fields relevant to the development of intelligence and humaneness.

Our goal is to foster a collaborative forum for researchers from various disciplines to interact, exchange ideas, and examine the implications of recent advancements in A.I. technology. Topics of discussion include but are not limited to: embodied A.I. and its potential to augment service robotics, A.I. alignment with human values and societal norms, the importance of autonomy in intelligent systems, the role of language in cognitive development, and the ethical, legal, and societal implications (ELSI) of A.I.

We believe that the challenges posed by the new generation of A.I. are unique and multifaceted, straddling the technical, the ethical, and the philosophical. From examining the neurophysiological underpinnings of cognitive development to employing predictive coding and the free-energy principle in world modeling, we aspire to delve into the heart of the matters that define our future with A.I.

The workshop also aspires to nurture a productive dialogue around topics like LLMs, foundation models for robotics, Metaverse for robots, affective computing, computational psychiatry, and the modeling of human cognition, each with potential applications in developing socially assistive robots that can better align with human needs.

Through this workshop, we hope to offer meaningful recommendations for future research directions, foster international collaborations, and lay the groundwork for the emergence of a symbiotic future between humans, robots, and A.I.

We welcome and encourage the participation of researchers interested in shaping this symbiotic future. We believe that the cross-fertilization of ideas will lead to the development of robots more attuned to human values, cognitive states, and societal needs, ultimately fostering a more integrated, harmonious co-existence between humans and A.I.