https://xitter.cf/t.co/CPLER1Iz1Z… | ICU consultant | affiliated associate professor, University of Cambridge | opinions my own

Cambridge, England
Joined October 2017
AriErcole retweeted
Epistemic Disturbance in the Graph Model for Conflict Resolution: State-Preserving Actions, Four-Valued Assessments, and the Distinction between Capability and Intention Yukiko Kato arxiv.org/abs/2610.11690 [𝚌𝚜.π™Άπšƒ 𝚌𝚜.𝙰𝙸 𝚌𝚜.𝙼𝙰]
2
9
528
Gaps in healthcare are everywhere, yet the 'institution' resists the meaningful change that #AI could bring, focusing instead on point-solutions that change very little. Why? What would it take to do better? My book now also as ePub... amzn.eu/d/0fFQNmCT
2
283
Much of how healthcare is constructed is a mess: #AI won't help that This 'manufactured complexity' stops improvement by making radical simplification almost impossible 𝘈𝘳𝘡π˜ͺ𝘧π˜ͺ𝘀π˜ͺ𝘒𝘭 𝘐𝘯𝘡𝘦𝘭𝘭π˜ͺ𝘨𝘦𝘯𝘀𝘦, π˜•π˜’π˜΅π˜Άπ˜³π˜’π˜­ π˜™π˜¦π˜΄π˜ͺ𝘴𝘡𝘒𝘯𝘀𝘦 ariercole.com/book
59
My book is out. π˜Όπ™§π™©π™žπ™›π™žπ™˜π™žπ™–π™‘ π™„π™£π™©π™šπ™‘π™‘π™žπ™œπ™šπ™£π™˜π™š, 𝙉𝙖𝙩π™ͺ𝙧𝙖𝙑 π™π™šπ™¨π™žπ™¨π™©π™–π™£π™˜π™šΒ is a book about why healthcare institutions absorb capable technology and produce almost no change amzn.eu/d/09KAhpIs
1
2
4
112
Who pays for open science? AI-ready datasets don't create themselves. Curating and maintaining research data is skilled work Data sharing can consume 20% of the budget If we want reproducible science & trustworthy AI, it needs funding @CC4AIM doi.org/10.1080/02699052.202…
103
We've spent 20 years asking the wrong question about NHS digital investment. "How many clicks will this save?" is reasonable. It's just not the right question.
1
2
1
281
Cross-sectional study, so correlation β‰  causation. Efficient trusts may become digitally mature, not the other way round. But the bigger issue is how we frame "digital" at all.
1
79
Health AI gets widely used but rarely transforms anything. Two reasons: The Partial Adoption Trap- widespread partial adoption is a stable equilibrium. The Value-Adoption Paradox- tools that transform care most are hardest to embed. arxiv.org/abs/2605.17388 @magdalenealumni
1
1
129
AriErcole retweeted
Incentives, Equilibria, and the Limits of Healthcare AI: A Game-Theoretic Perspective Ari Ercole arxiv.org/abs/2603.28825 [𝚌𝚜.π™Άπšƒ 𝚌𝚜.𝙰𝙸]
2
105
The President’s Commendation for a sustained contribution to the College was presented to Dr Rana Hafiz Ur Rehman. Dr Ari Ercole gave the Macintosh Professor Lecture. Seen here with President Dr Claire Shannon and Prof Iain Moppett. Read more: ow.ly/NsKo50UeHFx
1
8
13
2,227
AriErcole retweeted
To answer this, we learn from existing issues : 'what are the medico-legal implications if automated alerts are ignored & there is harm'.e.g., automated VTE alerts in EHR. @AriErcole ask a controversial question of we dont always get it right so does it matter if AI gets wrong
1
2
2
282
AriErcole retweeted
Now, I am at 'role of AI' in ICU in 2030. @ESICM #LIVES2024 @AriErcole Paul Elbers with Michel Van Genderen, Joanna Kloptowska, Massimiliano Greco, Anne Flint Data driven approach Level 1: data driven decision making all the way to Level 4: fully clinical decision support
1
7
13
1,814
Routinely collected data = huge opportunities for improving care "GOSLING" a tool to help researchers prevent re-identification Well done @DrAnmolArora ! doi.org/10.1371/journal.pone…
2
6
703
AriErcole retweeted
🎊 Wrapping up #ESICMData24, #Congratulations to the #TOP3 teams! 1️⃣ Clusterdocs (NL) 2️⃣ FAINT (BE) 3️⃣ ABO7 (Worldwide) πŸ‘Kudos to all participating teams, and faculty members for their dedicated contribution and unwavering commitment! 🌟 @drPaulElbers @AriErcole
1
2
1,335
Great work Chris and @tedinburgh_ on the epidemiology of sepsis in the AmsterdamUMCdb dataset. A must-read for any data scientist wanting to use this resource for sepsis research @drPaulElbers @patrickthoral @ESICM @CUH_NHS @MedCambridge @Cambridge_Uni doi.org/10.1371/journal.pone…
3
3
681