Embodied AI in Social Spaces: Responsible and Adaptive Robots in Complex Setting - UKAIRS 2025 (Copy)

Aleksandra Landowska, Aislinn Gomez Bergin, Ayodeji Opeyemi Abioye, Jayati Deshmukh, Andriana Bouadouki, Maria Wheadon, Athina Georgara, Dominic Price, Tuyen Nguyen, Shuang Ao, Lokesh Singh, Yi Long, Raffaele Miele, Joel E. Fischer, Sarvapali D. RamchurnarXiv 2025 · DOI · arXiv

TL;DR

This paper introduces a multidisciplinary project focused on developing responsible and adaptive AI-driven robots for complex social environments like care homes. By integrating co-design, ethical frameworks, and multimodal sensing, the project aims to create robots that are emotionally responsive, context-aware, and aligned with diverse user needs, ultimately fostering sustainable and human-centered futures.

What Problem Does This Paper Address?

The paper tackles the growing challenge of an aging global population and the resulting strain on healthcare and social care systems, particularly the shortage of human caregivers. While robots and AI offer potential solutions for assisting with tasks, companionship, and monitoring, their integration into care settings has been hindered by a narrow focus on technical performance, often neglecting the complex socio-technical, ethical, and emotional aspects of caregiving. This project aims to bridge this gap by developing robots that are not just functional but also trustworthy, transparent, empathetic, and ethically grounded, capable of adapting to individual needs and preferences within these sensitive environments.

Methods

The project employs a multidisciplinary approach, bringing together expertise in human-robot interaction, robotics, cognitive neuroscience, AI, ethics, and healthcare. The core methodology is structured around five work packages (WPs), with co-creation embedded throughout the entire project lifecycle. Key techniques include:

  • Co-design with stakeholders: Engaging older adults, carers, healthcare professionals, and ethicists to ensure the technology development is informed by real-world needs and values.
  • Multimodal sensing: Integrating physiological and behavioral sensors to understand human states and preferences.
  • Explainable adaptive AI: Developing AI models that can adapt robot behavior in real-time based on user input and context.
  • Responsible innovation principles: Embedding ethics, acceptability, regulatory compliance, and user empowerment from the outset.
  • Scenario prototyping and evaluation: Creating realistic care scenarios for testing and validating the systems.

Datasets

Not explicitly detailed in the provided text, but the paper mentions the creation of a "comprehensive multimodal dataset" integrating physiological and behavioral sensor data as a key output of Work Package 2.

Key Insights and Ideas

  • Beyond technical performance: The success of robots in social settings hinges on more than just their technical capabilities; ethical considerations, user acceptance, and emotional intelligence are paramount.
  • Co-creation is crucial: Involving all relevant stakeholders from the beginning is essential for developing technologies that truly meet user needs and integrate seamlessly into existing environments.
  • Embodied AI for empathy: Robots need to be "embodied" and capable of understanding and responding to human emotions and cognitive states to be truly effective and supportive in care settings.
  • Sustainable and human-centered futures: The project's ultimate goal is to leverage AI and robotics to enhance care quality and efficiency without replacing human interaction, thereby contributing to more sustainable and dignified care.

Key Results

The paper is an overview of a project in progress, so specific quantitative results are not yet available. However, the anticipated key outputs include:

  • Adaptive robotic systems trained using multimodal sensor data.
  • A multimodal dataset integrating neurophysiological and behavioral data.
  • AI models capable of interpreting user preferences, emotional states, and care-related requirements.
  • Realistic care scenarios and evaluation protocols.
  • Ethical frameworks and practical guidance for responsible deployment.
  • A demonstrator showcasing an adaptive and responsible multi-agent system.
  • Tools to assist care providers and policymakers in assessing and adopting responsible robotic solutions.

Limitations and Future Work

The paper itself is an overview of a project and does not explicitly detail limitations or future work beyond the scope of the project's planned outputs. However, the inherent complexity of developing truly adaptive and ethically sound AI for sensitive social environments implies significant challenges in achieving robust and universally applicable solutions. Future work will likely involve the rigorous evaluation and validation of the developed systems in real-world care settings, addressing any unforeseen ethical dilemmas, and refining the adaptive capabilities of the AI.

Summary updated

BibTeX
@misc{landowska2025embodied,
  author = {Aleksandra Landowska and Aislinn Gomez Bergin and Ayodeji Opeyemi Abioye and Jayati Deshmukh and Andriana Bouadouki and Maria Wheadon and Athina Georgara and Dominic Price and Tuyen Nguyen and Shuang Ao and Lokesh Singh and Yi Long and Raffaele Miele and Joel E. Fischer and Sarvapali D. Ramchurn},
  title = {{Embodied AI in Social Spaces: Responsible and Adaptive Robots in Complex Setting - UKAIRS 2025 (Copy)}},
  year = {2025},
  eprint = {2509.00218},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.00218},
}
Figure