Causality in Healthcare AI Hub
Lead Research Organisation:
University of Edinburgh
Department Name: Sch of Engineering
Abstract
The current AI paradigm at best reveals correlations between model input and output variables. This falls short of addressing health and healthcare challenges where knowing the causal relationship between interventions and outcomes is necessary and desirable. In addition, biases and vulnerability in AI systems arise, as models may pick up unwanted, spurious correlations from historic data, resulting in the widening of already existing health inequalities.
Causal AI is the key to unlock robust, responsible and trustworthy AI and transform challenging tasks such as early prediction, diagnosis and prevention of disease.
The Causality in Healthcare AI with Real Data (CHAI) Hub will bring together academia, industry, healthcare, and policy stakeholders to co-create the next-generation of world-leading artificial intelligence solutions that can predict outcomes of interventions and help choose personalised treatments, thus transforming health and healthcare. The CHAI Hub will develop novel methods to identify and account for causal relationships in complex data. The Hub will be built by the community for the community, amassing experts and stakeholders from across the UK to
1) push the boundaries of AI innovation;
2) develop cutting-edge solutions that drive desperately needed efficiency in resource-constrained healthcare systems; and
3) cement the UK's standing as a next-gen AI superpower.
The data complexity in heterogeneous and distributed environments such as healthcare exacerbates the risks of bias and vulnerability and introduces additional challenges that must be addressed. Modern clinical investigations need to mix structured and unstructured data sources (e.g. patient health records, and medical imaging exams) which current AI cannot integrate effectively. These gaps in current AI technology must be addressed in order to develop algorithms that can help to better understand disease mechanisms, predict outcomes and estimate the effects of treatments. This is important if we want to ensure the safe and responsible use of AI in personalised decision making.
Causal AI has the potential to unearth novel insights from observational data, formalise treatment effects, assess outcome likelihood, and estimate 'what-if' scenarios. Incorporating causal principles is critical for delivering on the National AI Strategy to ensure that AI is technically and clinically safe, transparent, fair and explainable.
The CHAI Hub will be formed by a founding consortium of powerhouses in AI, healthcare, and data science throughout the UK in a hub-spoke model with geographic reach and diversity. The hub will be based in Edinburgh's Bayes Centre (leveraging world-class expertise in AI, data-driven innovation in health applications, a robust health data ecosystem, entrepreneurship, and translation). Regional spokes will be in Manchester (expertise in both methods and translation of AI through the Institute for Data Science and AI, and Pankhurst Institute), London (hosted at KCL, representing also UCL and Imperial, leveraging London's rapidly growing AI ecosystem) and Exeter (leveraging strengths in philosophy of causal inference and ethics of AI).
The hub will develop a UK-wide multidisciplinary network for causal AI. Through extended collaborations with industry, policymakers and other stakeholders, we will expand the hub to deliver next-gen causal AI where it is needed most. We will work together to co-create, moving beyond co-ideation and co-design, to co-implementation, and co-evaluation where appropriate to ensure fit-for-purpose solutions
Our programme will be flexible, will embed trusted, responsible innovation and environmental sustainability considerations, will ensure that equality diversity and inclusion principles are reflected through all activities, and will ensure that knowledge generated through CHAI will continue to have real-world impact beyond the initial 60 months.
Causal AI is the key to unlock robust, responsible and trustworthy AI and transform challenging tasks such as early prediction, diagnosis and prevention of disease.
The Causality in Healthcare AI with Real Data (CHAI) Hub will bring together academia, industry, healthcare, and policy stakeholders to co-create the next-generation of world-leading artificial intelligence solutions that can predict outcomes of interventions and help choose personalised treatments, thus transforming health and healthcare. The CHAI Hub will develop novel methods to identify and account for causal relationships in complex data. The Hub will be built by the community for the community, amassing experts and stakeholders from across the UK to
1) push the boundaries of AI innovation;
2) develop cutting-edge solutions that drive desperately needed efficiency in resource-constrained healthcare systems; and
3) cement the UK's standing as a next-gen AI superpower.
The data complexity in heterogeneous and distributed environments such as healthcare exacerbates the risks of bias and vulnerability and introduces additional challenges that must be addressed. Modern clinical investigations need to mix structured and unstructured data sources (e.g. patient health records, and medical imaging exams) which current AI cannot integrate effectively. These gaps in current AI technology must be addressed in order to develop algorithms that can help to better understand disease mechanisms, predict outcomes and estimate the effects of treatments. This is important if we want to ensure the safe and responsible use of AI in personalised decision making.
Causal AI has the potential to unearth novel insights from observational data, formalise treatment effects, assess outcome likelihood, and estimate 'what-if' scenarios. Incorporating causal principles is critical for delivering on the National AI Strategy to ensure that AI is technically and clinically safe, transparent, fair and explainable.
The CHAI Hub will be formed by a founding consortium of powerhouses in AI, healthcare, and data science throughout the UK in a hub-spoke model with geographic reach and diversity. The hub will be based in Edinburgh's Bayes Centre (leveraging world-class expertise in AI, data-driven innovation in health applications, a robust health data ecosystem, entrepreneurship, and translation). Regional spokes will be in Manchester (expertise in both methods and translation of AI through the Institute for Data Science and AI, and Pankhurst Institute), London (hosted at KCL, representing also UCL and Imperial, leveraging London's rapidly growing AI ecosystem) and Exeter (leveraging strengths in philosophy of causal inference and ethics of AI).
The hub will develop a UK-wide multidisciplinary network for causal AI. Through extended collaborations with industry, policymakers and other stakeholders, we will expand the hub to deliver next-gen causal AI where it is needed most. We will work together to co-create, moving beyond co-ideation and co-design, to co-implementation, and co-evaluation where appropriate to ensure fit-for-purpose solutions
Our programme will be flexible, will embed trusted, responsible innovation and environmental sustainability considerations, will ensure that equality diversity and inclusion principles are reflected through all activities, and will ensure that knowledge generated through CHAI will continue to have real-world impact beyond the initial 60 months.
Organisations
- University of Edinburgh (Lead Research Organisation, Project Partner)
- Canon (Collaboration)
- Life Sciences Scotland (Project Partner)
- Spectra Analytics (Project Partner)
- Institute of Cancer Research (Project Partner)
- Chief Scientist Office (CSO), Scotland (Project Partner)
- University of California Berkeley (Project Partner)
- Zeit Medical (Project Partner)
- The Data Lab (Project Partner)
- Healthcare Improvement Scotland (Project Partner)
- PrecisionLife Ltd (Project Partner)
- Microsoft Research Limited (Project Partner)
- Samsung AI Centre (Project Partner)
- CANCER RESEARCH UK (Project Partner)
- Kheiron Medical Technologies (Project Partner)
- National Institute for Health and Care Excellence (NICE) (Project Partner)
- The MathWorks Inc (Project Partner)
- McGill University (Project Partner)
- Willows Health (Project Partner)
- Queen Mary University of London (Project Partner)
- ARCHIMEDES (Project Partner)
- Hurdle (Project Partner)
- UCL Hospitals NHS Foundation Trust (Project Partner)
- Indiana University Bloomington (Project Partner)
- Scotland 5G Centre (Project Partner)
- Evergreen Life (Project Partner)
- NHS Lothian (Project Partner)
- Data Science for Health Equity (Project Partner)
- Manchester Cancer Research Centre (Project Partner)
- UCB Pharma (United Kingdom) (Project Partner)
- Health Data Research UK (Project Partner)
- Mayo Clinic and Foundation (Rochester) (Project Partner)
- Amazon Web Services (Not UK) (Project Partner)
- Huawei Technologies (UK) Co Ltd (Project Partner)
- Gendius Limited (Project Partner)
- Research Data Scotland (Project Partner)
- NHS NATIONAL SERVICES SCOTLAND (Project Partner)
- DHI: Digital Health & Care Innovation Centre (Project Partner)
- Bering Limited (Project Partner)
- Meta (Previously Facebook) (Project Partner)
- Endeavour Health Charitable Trust (Project Partner)
- CausaLens (Project Partner)
- Canon Medical Research Europe Ltd (Project Partner)
- NHS GREATER GLASGOW AND CLYDE (Project Partner)
- British Standards Institution BSI (Project Partner)
- Scottish AI Alliance (Project Partner)
- Scottish Ambulance Service (Project Partner)
- University of Dundee (Project Partner)
- Sibel Health (Project Partner)
- ELLIS (Project Partner)
Publications
Charles Jones
(2025)
Rethinking Fair Representation Learning for Performance-Sensitive Tasks
Chockler H
(2024)
Explaining Image Classifiers
Chockler H.
(2024)
Explaining Image Classifiers
in Proceedings of the International Conference on Knowledge Representation and Reasoning
David Watson
(2024)
Bounding Causal Effects with Leaky Instruments
in Bounding Causal Effects with Leaky Instruments
Hana Chockler
(2024)
A Causal Analysis of Harm Sander Beckers, Hana Chockler, Joseph Y. Halpern Minds and Machines (2024) 34:34
in A Causal Analysis of Harm
Ilaria Pia La Torre
(2025)
To BEE or not to BEE: Estimating more than Entropy with Biased Entropy Estimators
J Penn, LM Gunderson, G Bravo-Hermsdorff, R Silva, DS Watson
(2024)
BudgetIV: Optimal Partial Identification of Causal Effects with Mostly Invalid Instruments
Jialin Yu
(2024)
Structured Learning of Compositional Sequential Interventions
| Description | FUTURE-AI: international consensus guideline for trustworthy and deployable AI in healthcare |
| Geographic Reach | Multiple continents/international |
| Policy Influence Type | Participation in a guidance/advisory committee |
| URL | https://future-ai.eu/ |
| Description | Presentation at the Life Sciences Scotland Industry Leadership Group at Scottish Parliament |
| Geographic Reach | National |
| Policy Influence Type | Participation in a guidance/advisory committee |
| Description | STANDING Together: STANdards for data Diversity, INclusivity and Generalisability |
| Geographic Reach | Multiple continents/international |
| Policy Influence Type | Participation in a guidance/advisory committee |
| Impact | STANDING Together recommendations are highlighted by regulators in multiple countries as best practices. |
| URL | https://www.datadiversity.org/ |
| Description | The Regulatory Horizons Council event |
| Geographic Reach | National |
| Policy Influence Type | Contribution to a national consultation/review |
| Title | Benchmarking Counterfactual Image Generation |
| Description | This repository contains the code for the paper "Benchmarking Counterfactual Image Generation" |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2024 |
| Provided To Others? | Yes |
| Impact | Nothing to report yet. |
| URL | https://gulnazaki.github.io/counterfactual-benchmark/ |
| Title | CHARIOT - cardiovascular prevention causal prediction model |
| Description | This tool allows individuals to estimate their future risk of cardiovascular events under interventions. Currently statins and antihypertensives are supported - lifestyle interventions will be added soon. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | In discussions with industrial partners regarding evaluating and implementing the approaches in patient facing health records, and clinical practice. |
| URL | https://alexpate30.shinyapps.io/chariot_prototype_shiny2/ |
| Title | Counterfactual contrastive learning |
| Description | This repository contains the code for the papers "Counterfactual contrastive learning: robust representations via causal image synthesis" and extended version "Robust representations for image classification via counterfactual contrastive learning". |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2024 |
| Provided To Others? | Yes |
| Impact | Nothing to report yet. |
| URL | https://github.com/biomedia-mira/counterfactual-contrastive |
| Title | Dual Risk Minimization |
| Description | A method for learning and performing automated classification across data sources of varying relationships, while trading-off robustness to major unanticipated variabilities against making the most of the available data. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2024 |
| Provided To Others? | Yes |
| Impact | Nothing to report yet. |
| URL | https://github.com/vaynexie/DRM |
| Title | Structured Learning of Compositional Sequential Interventions |
| Description | Code for a predictive model of effects of combinations of interventions over time in sequential data. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2024 |
| Provided To Others? | Yes |
| Impact | Many organisations track progress of units over time, and expose them to interventions intended to guide their behaviour. We have described the ideas and outcomes of this projects to collaborators in Spotify who were acknowledged in the companion paper. |
| URL | https://github.com/jialin-yu/CSI-VAE |
| Description | CHAI-Canon Medical Research Europe |
| Organisation | Canon |
| Department | Canon Medical Research Europe |
| Country | United Kingdom |
| Sector | Academic/University |
| PI Contribution | The partnership is in an early-stage. Two PhD Studentships are in scope, which are currently being recruited for. The focus is causal AI for medical imaging. |
| Collaborator Contribution | To early to specify. |
| Impact | Not yet relevant. |
| Start Year | 2025 |
| Description | Causal AI in Healthcare workshop |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Professional Practitioners |
| Results and Impact | In collaboration with the Exeter Health Analytics research network we organized a workshop to introduce the local community of the University of Exeter to the CHAI Hub. The event included talks by co-Is Sperrin (Manchester), Silva (UCL), Li (ICL), and grant application runway activity. Multiple teams have since reported that they are planning to apply to CHAI's pump-priming fund. |
| Year(s) Of Engagement Activity | 2025 |
| URL | https://www.chai.ac.uk/events |
| Description | Causality in Science and AI meeting series |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Other audiences |
| Results and Impact | In collaboration with the Institute for Data Science and Artificial Intelligence of the University of Exeter, and the Egenis Research Centre, we organized a five meeting series on the topic of Causality in Science and AI, and including a talk by Dr Sander Beckers (Cornell). The series has drawn audience from heterogeneous networks and disciplines, and including academics connecting remotely from the EU. |
| Year(s) Of Engagement Activity | 2025 |
| URL | https://www.exeter.ac.uk/research/centres/egenis/activities/specialinterestgroups/#a8 |
| Description | Panel Discussion - Pioneering AI in multiple long term conditions conference, Sep 2024 |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Patients, carers and/or patient groups |
| Results and Impact | Contributed a brief presentation and was a panel member in discussion 'Can AI, data science & digital innovation transform MLTC Health & Social Care research?' There was a broad mix of patients, practitioners, researchers and industry in attendance. |
| Year(s) Of Engagement Activity | 2024 |
| URL | https://www.turing.ac.uk/events/pioneering-ai-mltc-bridging-research-and-practice-conference-2024 |
| Description | Panel discussion (RCR-NHS Global AI Conference) |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Professional Practitioners |
| Results and Impact | Ben Glocker chaired a session and panel discussion on the topic of 'Transparency and bias' under the 'Stream 3: Governance and regulation' at the RCR-NHS Global AI Conference 2025. |
| Year(s) Of Engagement Activity | 2025 |
| URL | https://rcraiconference.com/2025/pages/stream3_session2 |
| Description | Panel member, Causal Representation Learning workshop |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Professional Practitioners |
| Results and Impact | A panel discussion among experts on the state and future directions of a major area within causal models in AI. Audience participation and questions helped to frame possible research programmes. |
| Year(s) Of Engagement Activity | 2024 |
| URL | https://crl-community.github.io/neurips24 |
| Description | Pixel Pandemonium at the European Congress of Radiology 2025 |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Professional Practitioners |
| Results and Impact | This tech demo showcases cutting-edge research in generative AI for high-fidelity medical image synthesis. This interactive demo allows users to explore AI-generated counterfactual images, including chest radiographs and mammograms. With a press of a button, users can simulate 'what-if' scenarios: How would this patient's scan look like with and without disease? How would the image characteristics change if it had been acquired with a different scanner? How would this mammogram look like if the breast tissue had a different density? These imaginable images have potential use for stress-testing AI models and to improve their robustness, reliability and fairness. |
| Year(s) Of Engagement Activity | 2025 |
| URL | https://connect.myesr.org/ecr-2025/pixel-pandemonium/?tabref=1 |
| Description | Seminar talk, University of Toronto |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Postgraduate students |
| Results and Impact | A one-day visit to University of Toronto as their first external speaker on the series "Causal Inference: Bringing together data science and causal inference for better policy recommendations". Had several individual face-to-face meeting with faculty from Statistics, Computer Science, Economics and Business from UofT. Made new connections for CHAI from researchers on AI in healthcare. |
| Year(s) Of Engagement Activity | 2024 |
| URL | https://datasciences.utoronto.ca/causal_inference/ |