www.sciencedirect.com/science/arti... Behind this paper: a physician from Iran, a social scientist from Australia, a behavioral scientist from Norway, a history and philosophy enthusiast from the US, and a data scientist from the US.

Epistemic humility for physicians and scientists

sciencedirect.com

To ensure AI delivers meaningful value, we need five fundamental tools: a mirror, a flashlight, a microscope, a paintbrush, and a podium. The mirror helps us examine our own motivations and values—ensuring we're building AI for the right reasons.

www.nature.com/articles/s41... We need AI that will keep humans on the rails, and humans that will keep AI on the rails. Bruno Latour would say AI systems aren't neutral tools but active participants in networks that reconfigure relationships between humans, institutions, and knowledge production.

Teaching machines to doubt - Nature Medicine

We must build safeguards against AI’s most dangerous feature: its ability to stop us thinking critically.

nature.com

datascienceandhealth.ubc.ca/events/ai-fu... Join us at the AI & Future of Medicine at the University of British Columbia in Vancouver, Canada on November 15-16, 2025 as we forge a path forward in the chaos that the AI hype has wrought.

Medical education does not exist in a vacuum, and attempts to reform it in isolation are fundamentally inadequate if the broader healthcare system remains broken. (The illustration is provided by Dancing with Markers.)

sites.google.com/view/ai-fait... Please join us at the Vatican on September 8th for dialogues around AI and faith. "Faith and Artificial Intelligence: Using language models to promote human connection“

AI & Faith Dialogues

AI & Faith Dialogues

sites.google.com

arxiv.org/abs/2508.16519 How do we make AI worth its cost? Simple: We fix the world. First up: Publish or perish culture.

The Community Index: A More Comprehensive Approach to Assessing Scholarly Impact

The h index is a widely recognized metric for assessing the research impact of scholars, defined as the maximum value h such that the scholar has published h papers each cited at least h times. While ...

arxiv.org

www.techrxiv.org/users/953594... We are at a critical juncture where AI could either extend scientific progress or flood the literature with confident mirages—scientific deepfakes that erode the trust science needs to correct itself.

open.spotify.com/show/7w2YAD1... For those who were not able to attend last Monday's event on the bias of language and how LLMs will shape the course of society, here is a link to the audio recordings of the talks and panel discussions.

arxiv.org/abs/2508.00952 Academic laboratories face mounting resource constraints: budgets are tightening, grant overheads are potentially being capped, and the market rate for data-science talent significantly outstrips university compensation.

arxiv.org/abs/2508.00952

Academic Vibe Coding: Opportunities for Accelerating Research in an Era of Resource Constraint

Academic laboratories face mounting resource constraints: budgets are tightening, grant overheads are potentially being capped, and the market rate for data-science talent significantly outstrips univ...

arxiv.org

forms.cloud.microsoft/pages/respon... Please join us at the National University of Singapore (NUS) Seminar for Responsible AI for Eldercare on August 27th from 14.00 - 16.30 at MD1, Tahir Foundation Building. The global population is aging while the healthcare workforce is shrinking.

sites.google.com/view/mithic-... A meeting of the hearts and souls at MIT. Please join us on August 18th (Monday) for a day of reflection in this age of AI. MIT Critical Data and MIT Global Languages partnered to explore the bias of language itself and how language models might shape human behavior.

MITHIC Datathon

sites.google.com

elty.it/medico/hacka... Please join us at Google Italy office in Milan on September 3-4 as we explore foundational issues with AI in the health space through hands-on workshops on data interrogation and critical thinking.

www.biomedcentral.com/collections/... Artificial intelligence is not about coding or developing algorithms. It's about critical thinking and reimagining systems equipped with powerful tools that retrieve information, create and validate knowledge, call out biases, and challenge the statusquo

Bangkok datathon mentors July 2025. Only 3 mentors including myself were not local. This is the only sustainable and scalable strategy. AI by us for all of us. The students from different schools planned the entire event.

informatics.bmj.com/content/32/1... To everyone building AI models, please don't stop when you have a model that performs accurately on an unseen test set. You need to figure out whether the model uses "shortcut" features.

Bias in vital signs? Machine learning models can learn patients’ race or ethnicity from the values of vital signs alone

Objectives To investigate whether machine learning (ML) algorithms can learn racial or ethnic information from the vital signs alone.Methods A retrospective cohort study of critically ill patients bet...

informatics.bmj.com

news.mit.edu/2025/changin... AI is our chance to rewrite the rules. If we fail, it’s because we failed to dream big enough about how a reimagined world could look.

Changing the conversation in health care

The Language/AI Incubator, a cross-disciplinary, MITHIC-funded project, investigates how artificial intelligence can improve communications among patients and practitioners.

news.mit.edu

www.sciencedirect.com/science/arti... AI technologies contribute to climate change and sociopolitical instability due to their high consumption of natural, energy and human resources that are typically "extracted" within power structures.

Climate change and health: the next challenge of ethical AI

Artificial intelligence (AI) is one of the world's most resource-intensive digital technologies, but the environmental impact of AI on health remains …

sciencedirect.com

journals.lww.com/ccejournal/f... In this paper, we investigated the model performance of GOSSIS (Global Open Source Severity of Illness Score) under geographical, temporal and pandemic dataset shift. AI Models should be bound by space and time. In other words, AI models have expiration dates.

Mortality Prediction Performance Under Geographical,... : Critical Care Explorations

aluation in line with contemporary healthcare settings and regional considerations. OBJECTIVES: To evaluate the geographic and temporal generalizability of the models for mortality prediction in IC...

journals.lww.com

dl.acm.org/doi/10.1145/... Machine learning courses are not teaching students how to understand the data. Everyone is racing just to build the next model with the best accuracy based on very flawed data. AI is tempting us with shortcuts to traditional rewards such as papers and citations.

Race Against the Machine Learning Courses | ACM Transactions on Intelligent Systems and Technology

Despite the rapid integration of AI in healthcare, a critical gap exists in current machine learning courses: the lack of education on identifying and mitigating bias in datasets. This oversight risks perpetuating existing health disparities through ...

dl.acm.org

www.sciencedirect.com/science/arti... It is crucial that AI developers understand the "backstory" of the data. Who made it to the cohort, and who did not? Sampling selection bias leads to spurious associations and/or shortcut features that are easily learned by algorithms.

The effect of selection bias on the performance of a deep learning-based intraoperative hypotension prediction model using real-world samples from a publicly available database

There are models to predict intraoperative hypotension from arterial pressure waveforms. Selection bias in datasets used for model development and val…

sciencedirect.com

link.springer.com/article/10.1... It is alluring to think that AI will deliver care to the 4.5 billion without access to a health system. But "generating workflows that require building trust and relationships" cannot be conjured by AI.

Building health systems capable of leveraging AI: applying Paul Farmer’s 5S framework for equitable global health - BMC Global and Public Health

The development of artificial intelligence (AI) applications in healthcare is often positioned as a solution to the greatest challenges facing global health. Advocates propose that AI can bridge gaps ...

link.springer.com

www.sciencedirect.com/science/arti... If your business model is still selling algorithms, please stop. The accuracy of models is limited in time and space. The pursuit of generalizability is misguided. AI should be local, iterative, reflective and reflexive.

Global disparities in artificial intelligence-based mammogram interpretation for breast cancer: A scientometric analysis of representation, trends, and equity

Breast cancer (BC) is the most frequently diagnosed cancer and the leading cause of cancer death among women worldwide. Artificial intelligence (AI) s…

sciencedirect.com

www.biomedcentral.com/collections/... We are still accepting submissions to a special collection around AI for the journal Critical Care.We are not interested in papers that are only developing algorithms with near-perfect accuracy.We are already drowning in models. We are particularly interested in

www.medrxiv.org/content/10.1... This paper is published today behind a paywall so I'm sharing the preprint here. It describes a causal inference framework to evaluate impact of life-sustaining therapies in the ICU across different demographic axes.

Evaluating equitable care in the ICU: Creating a causal inference framework to assess the impact of life-sustaining interventions across racial and ethnic groups

Background Variability in the provision of intensive care unit (ICU)-interventions may lead to disparities between socially defined racial-ethnic groups. Research Question We used causal inference to...

medrxiv.org

www.sciencedirect.com/science/arti... The collection of extensive demographic data raises important concerns that include the construction of intersectional social categories (ie, race and its shifting meaning in different sociopolitical contexts), the risks of biological reductionism, (cont.)

Weighing the benefits and risks of collecting race and ethnicity data in clinical settings for medical artificial intelligence

Many countries around the world do not collect race and ethnicity data in clinical settings. Without such identified data, it is difficult to identify…

sciencedirect.com

www.atsjournals.org/doi/10.1164/... Not only are the medical devices and biomarkers not optimized across patient demographics and other features, the intensity/frequency of measurement varies across patients.

www.jscai.org/article/S277... Knowledge, our understanding of truth, is purely based on experiences and observations, and that there is no such thing as “ground truth”. How we see things objectifies, or more aptly, “subjectifies” those things.

journals.plos.org/digitalhealt... The pursuit of diversity goes much deeper than simply a core tenet of some "liberal" ideology. For artificial intelligence, specifically reinforcement learning, successful adaptation to changing conditions requires a level of diversity among agents' actions.

Introducing the Team Card: Enhancing governance for medical Artificial Intelligence (AI) systems in the age of complexity

Author summary Bias in medical artificial intelligence (AI) systems refers to systematic errors that unfairly favor or disadvantage certain groups or outcomes. While often attributed to data quality i...

journals.plos.org