Dr Cristina Palmero Cantarino - King’s College London

Research Fellowships 2025

Population ageing will soon exert significant pressure on healthcare systems, economic productivity, and society at large, demanding immediate and targeted solutions. As a result, the demand for embodied social AI systems is expected to surge in the upcoming years to streamline care and personal support in our homes and workplaces. Seamless human-machine interaction requires sophisticated yet transparent technologies that accurately perceive and interpret our surroundings. This is crucial for ensuring interactions are safe, timely, and aligned with user needs, directly impacting widespread adoption. Current technologies can detect human actions and the physical world effectively through various sensors, but still struggle with social contexts and the complex, dynamic, and ambiguous nature of human behaviour. Behaviour is influenced by an interplay of socio-cultural, situational, personal, and interpersonal factors. Comprehensive understanding requires perceiving and reasoning about these factors explicitly and jointly an area where existing technologies, often reliant on purely data-driven machine learning approaches, fall short.

The scarcity of real-world in-person sensory data further limits robustness and generalisation across contexts. ECASIS aims to address these unmet needs with a novel computational framework for holistic, context-aware social AI. The project proposes three integrated advances: 1) augmenting sensory data with extensive behavioural sciences knowledge; 2) merging statistical learning with explicit reasoning through human-readable behaviour and context concepts; and 3) enabling understanding of multiple factors jointly. This hybrid approach will allow systems to make informed and explainable inferences, akin to human cognition. ECASIS will follow an ethics-first interdisciplinary approach through close academic-industry collaboration, incorporating insights from computer vision, neuro-symbolic computing, responsible and social AI, and behavioural sciences. Its real-world impact will be showcased through applications like socially aware robot navigation and personalised well-being systems. Additionally, ECASIS will provide open-source toolboxes and new benchmarks, driving innovation and collaboration to enhance quality of life through advanced social AI. 

Personal website

LinkedIn

X (Twitter)