How to develop AI that addresses health inequities

Joe Alderman has this great ability to break down complex things into a good hearty discussion. Joe is an Anaesthetist by night and AI academic by day. His insights on what it takes to deploy and monitor AI in healthcare with a lens on not leaving people behind are relevant wherever in the world you are.

We discussed some of the nuance and challenges around AI in healthcare, as well as some anti patterns 

• If you are a data or AI scientist, or shipping AI products, what can you do starting tomorrow to begin thinking more proactively about underserved communities or ensure you’re not creating unintended harm to vulnerable groups. 
• Stuff the industry is not doing great on: We’re starting with the dataset or the technology we happen to have, instead of working backwards from what we are fundamentally trying to achieve in the world. (If I nodded any more at this point my head would have rolled off)
• A revisit on where things are with the Standing Together initiative, something Xiao Liu talked about in the very first episode.  In the UK, there are big areas where life expectancy is falling. Health inequalities are rising. Our datasets have bias. It’s tempting to fall into the narrative that this is a tech deficiency problem. It is not. It is a whole society problem and we’re not going to solve it with JUST tech. 
• If govts, regulators and health systems find it challenging to be on top of the safety issues with LLMs, how are lay people going to? Joe covers how the upcoming work of the Healthcare Chatbot User Guide can help people navigate some of the key things they need to consider

We cover a LOT in this 1 hour discussion, including how patients and clinicians can think more critically about which tasks to best use Gen AI with, and how to think about safety and liability. 

This one is for AI and data scientists, product folk, developers and clinicians. Get it on your weekend playlist

Links:

MHRA’s National Commission into the Regulation of AI in Healthcare: https://www.gov.uk/government/groups/national-commission-into-the-regulation-of-ai-in-healthcare

Building The Health Chatbot Users’ Guide: https://healthchatbotguide.org/

The NHS Fellowship in Clinical AI: https://www.nhsfellowship.ai/

NHS Digital Clinical Safety Training: https://digital.nhs.uk/services/clinical-safety/clinical-risk-management-training

Chapters
00:00 Introduction and Joe's work
06:36 The Role of AI in Patient Care
07:52 Navigating AI Risks and Benefits
18:13 Addressing Health Data Poverty
24:28 Building Responsible AI Systems
30:18 Exploring Large Language Models in Healthcare
34:05 Addressing Bias and Representativeness in AI Models
40:04 Navigating Regulatory Challenges for AI Technologies
40:51 Empowering Patients in the Age of AI
44:32 Public Perception and Trust in Digital Health
47:00 Identifying Real Problems in Healthcare Technology
52:00 The Future of AI in Healthcare
55:29 Joe's insights for those building in low resource settings
56:57 A book recommendation and critical thinking for patients and clinicians

About Joe

Joe is an NHS anaesthetist in the UK and an NIHR clinical lecturer (assistant professor) in AI and digital health at the University of Birmingham, UK. 

He leads mixed-methods research and policy engagement to help get the most from AI in healthcare - including the STANDING Together initiative. He recently founded an international initiative to build ‘The Health Chatbot Users’ Guide’.