CHARM
CHecklist for
Actioning
Responsible
MLOps
In the near future, clinical artificial intelligence will become an integral part of patient care pathways. The CHARM international consensus group are developing process guidelines to help healthcare organisations deploy, maintain, and scale AI. We are committed to improving safety, quality, and equity for patients around the world, through responsible AI implementation.
Why CHARM?
All too often, healthcare AI deployments are 'fire and forget', conveniently ignoring negative biases in clinical data and algorithms, and the dynamic nature of patient populations. The need for responsible AI is well understood, but this can be difficult to action for healthcare organisations around the world - especially those that are at the start of an AI journey. CHARM turns concepts for responsible healthcare AI into step-by-step items on a practical checklist, ensuring that all important considerations are covered. These support an MLOps lifecycle: the continuous processes that maintain safety and effectiveness, and reinforce trust for patients and clinicians.
Concept to Process
Building on key concepts for responsible AI for healthcare, CHARM will translate critical components into a practical checklist that healthcare organisations can easily follow
ACTIONABLE PROCESSES
CHARM sits between conceptual responsible AI frameworks, and prescriptive technical MLOps tools. Each CHARM item will be a recommended process that can help healthcare organisations manage AI model deployments.
CRITICAL ELEMENTS
Processes are expected to be critical to the responsible use of AI in patient pathways. That is to say, their exclusion from AI deployment practices may risk safety, effectiveness, quality of care, and patient/clinician trust.
GLOBAL &
INCLUSIVE
Most of the world is only just beginning their healthcare AI journey. Our priority is inclusivity of global and low resource regions, where AI is poised for transformative impact, and where regulation may lag behind.
The CHARM project team is based out of the University of Oxford, Imperial College London, and the Massachusetts Institute of Technology. The principal investigators are funded by the Wellcome Trust. While we collaborate with a wide range of cross-sector stakeholders, we receive no industry sponsorship or funding.