Artificial intelligence is increasingly being used across healthcare and social care to support prediction, information retrieval, documentation, monitoring and decision-making. However, many important care decisions require more than identifying patterns in data. They require understanding why an outcome occurred, what factors contributed to it, and what could have been done differently. Causal AI offers an important opportunity to move beyond prediction towards systems that can help reason about causes, interventions and possible outcomes. This one-day interdisciplinary workshop will bring together researchers in causal AI, responsible AI and healthcare technologies with healthcare and social care professionals, care providers, technology developers, NHS and public-sector organisations, policymakers and people with lived experience.

​

Particular attention will be given to agentic AI: AI systems capable of carrying out multiple steps such as retrieving information, identifying missing evidence, summarising records and supporting investigations. Their use in sensitive care environments raises important questions around causality, uncertainty, human oversight, accountability and safe escalation. The workshop will explore where causal AI can provide genuine value, what evidence and data are required, what barriers may prevent adoption, and how researchers, healthcare and social care organisations and technology providers can work together to develop trustworthy solutions.

​

Why causal AI matters:  From identifying patterns to examining possible interventions

​

Predictive AI can identify patterns associated with an outcome. Causal reasoning examines how factors may influence that outcome and what might happen if an intervention changes them.

​

In care settings, this raises questions such as:

​

• Which factors may have contributed to a fall, medication-related incident or missed visit?

• What evidence is missing before a conclusion can be reached?

• Could an observed relationship be explained by another factor?

• What intervention might have prevented the outcome?

• When should an AI system request more information or defer to a professional?

• ​

The workshop will explore these questions while recognising uncertainty, evidence gaps and the limits of AI-supported reasoning.