Research
Research Interests
Decision-theoretic and responsible artificial intelligence (AI), cognitive modelling and causal reasoning, transfer and reinforcement learning, human-aware and adaptive computing, and biomedical and health informatics.
Research Activities
I direct a multi-disciplinary research and innovation agenda on human-aware decision modelling in complex environments, mixed-initiative learning from multiple knowledge sources, , and responsible AI for changing ecosystems.
My research aims at understanding brain, mind, and computation functions to develop responsible AI techniques, systems, and solutions that will improve quality of life and increase happiness for humankind.
My team includes academics, researchers, and students at NUS, with collaborators from local and overseas institutions and hospitals. Besides publishing in international journals and conferences, we have commercialized some of the technologies, and published open-source tools and systems that are internationally distributed. We are also actively involved in national and international scientific advisory and policy guidance work related to the ethical, governance, and regulatory considerations for AI, especially for AI in Health.
We carry out our work in:
NUS Artificial Intelligence Lab
Medical Computing Lab
NUS Advanced Robotics Centre
Research Directions
My team’s recent work focuses on representation change, causal reasoning, multimodal learning, human-guided reinforcement and transfer learning, and responsible decision making in rapid changing environments.
• The technical challenges include handling heterogeneous, noisy, and sparse information, and representation and reasoning with change, with emphases on managing complexity, uncertainty, and dynamicity in the world.
• The system challenges include designing adaptive systems that will evolve with changes in the technical functionalities, system infrastructures, usage patterns, and operational contexts.
Our research anchors on the Learning from Real-World Information and Reasoning with Change themes, to support responsible, adaptive decision making in complex, rapidly changing environments. Our technical approach is based on Bayesian theory, decision theory, statistical learning, causal reasoning, and social cognitive neuroscience that serves as a unifying methodology foundation, motivated by and tested in a wide range of real-world settings.
The target applications include prediction and decision analytics, human-aware robotics, game artificial intelligence, personalized and lifelong education, and assistive care for elderly with neurocognitive disorders, through the continuum of care.
The main challenges include:
• Reasoning at multiple level of details and targeting different users and perspectives;
• Learning with sparse, multi-modal information from multiple, heterogeneous sources;
• Decision making with varying human-centric motivations, objectives, beliefs, and values
The main topics include:
• Adaptive representation and reasoning
• Natural intelligence inspired representation discovery and adaptation
• Active, continual, and transfer learning in temporal causal models
• Integrated learning from human knowledge and real-world data
• Responsible, multi-modal learning from multiple sources
• Causal inference from multiple information sources
• Context-sensitive decision making
• Human guided, hierarchical reinforcement learning
• Mixed-initiative, responsible dynamic decision making
The main domains include:
• Biomedical and Health informatics
• Game Artificial Intelligence, assistive robotics, and personalized, lifelong education.