In today’s Duke AI Health Friday Roundup: White House OSTP releases “AI Bill of Rights”; Pääbo wins Physiology/Medicine Nobel for paleogenomics; COVID lessons and coming pandemics; messaging as public health tool; postdoc pipeline slows to a trickle; systematic review finds paucity of randomized trials of machine learning interventions; the limits of mental health chatbots; common pitfalls of AI journalism, much more:
AI, STATISTICS & DATA SCIENCE
- “…despite the large number of medical machine learning–based algorithms in development, few [randomized clinical trials] RCTs for these technologies have been conducted. Among published RCTs, there was high variability in adherence to reporting standards and risk of bias and a lack of participants from underrepresented minority groups.” A systematic review, published by Plana and colleagues in JAMA Network Open, examines the state of randomized research evaluating machine-learning based healthcare interventions.
- “In America and around the world, systems supposed to help with patient care have proven unsafe, ineffective, or biased. Algorithms used in hiring and credit decisions have been found to reflect and reproduce existing unwanted inequities or embed new harmful bias and discrimination. Unchecked social media data collection has been used to threaten people’s opportunities, undermine their privacy, or pervasively track their activity—often without their knowledge or consent.” Big news in the world of AI: the White House’s Office of Science and Technology Policy has released a “Blueprint for an AI Bill of Rights” this week. (For comparison, consider also the European Union’s recently released text for its own proposed Artificial Intelligence Act).
- “We noticed that many articles tend to mislead in similar ways, so we analyzed over 50 articles about AI from major publications, from which we compiled 18 recurring pitfalls. We hope that being familiar with these will help you detect hype whenever you see it. We also hope this compilation of pitfalls will help journalists avoid them.” At the AI Snake Oil Substack, Sayash Kapoor and Arvind Narayanan describe some of the most common forms of hype and exaggeration that afflict journalism about AI, and offer tips for countering it.
- “Our agent, AlphaTensor, is trained to play a single-player game where the objective is finding tensor decompositions within a finite factor space. AlphaTensor discovered algorithms that outperform the state-of-the-art complexity for many matrix sizes. Particularly relevant is the case of 4 × 4 matrices in a finite field, where AlphaTensor’s algorithm improves on Strassen’s two-level algorithm for the first time, to our knowledge, since its discovery 50 years ago…”A research article published this week in Nature by researchers from the London DeepMind project describes an AI agent trained to discover high-efficiency algorithms for matrix multiplication.
- “In the field of mental health, few new areas generate as much excitement as machine learning, which uses computer algorithms to better predict human behavior. There is, at the same time, exploding interest in biosensors that can track a person’s mood in real time, factoring in music choices, social media posts, facial expression and vocal expression.” The New York Times’ Ellen Barry reports on an ongoing project that is exploring whether individual biometric data gathered from smartphones can be used to predict suicide risk.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “American leaders and pundits have been trying to call an end to the pandemic since its beginning, only to be faced with new surges or variants. This mindset not only compromises the nation’s ability to manage COVID, but also leaves it vulnerable to other outbreaks. Future pandemics aren’t hypothetical; they’re inevitable and imminent. New infectious diseases have regularly emerged throughout recent decades, and climate change is quickening the pace of such events. As rising temperatures force animals to relocate, species that have never coexisted will meet, allowing the viruses within them to find new hosts—humans included. Dealing with all of this again is a matter of when, not if.” At The Atlantic, Ed Yong offers a sobering appraisal of US readiness to meet a future pandemic, even given the lessons of the past few years.
- “…using sophisticated DNA sequencing methods, he and his colleagues went on to sequence the full Neandertal genome, publishing their findings in 2010. The team found that the most recent common ancestor of Homo sapiens and Neandertals lived about 800,000 years ago, and that the two species interbred over thousands of years. About 1 to 4 percent of the genomes of modern humans of European or Asian descent comes from Neandertals.” Scientific American’s Tanya Lewis reports that paleogeneticist Svante Pääbo has been awarded the 2022 Nobel Prize in Physiology or Medicine for his work decoding the genomes of ancient hominin lineages.
- “Smoking causes substantial economic loss in the USA. Tobacco control efforts that lower the prevalence of smoking equitably can contribute considerably to improved macroeconomic performance in the short and long term by reducing health expenditures and avoiding productivity losses.” A research article published by Nargis and colleagues in Lancet Public Health reckons up the economic loss stemming from cigarette smoking in the US.
- “Beyond the research community, the Dashboard has been relied on by all levels of government and commercial entities in multiple countries, states/provinces, and cities around the world for informing COVID-19 response. As examples, the US Centers for Disease Control and Prevention (CDC), Department of Health and Human Services, and Federal Emergency Management Agency used the JHU data to guide US policy, Johnson & Johnson relied on the data to determine where best to hold vaccine trials, and Ford Motor Company used the data to determine which factories and offices to keep operational and to time the reopening of others.” At JAMA, Lasker-Bloomberg Public Service Award winner Lauren Gardner describes the COVID tracking dashboard she and colleagues at Johns Hopkins developed to provide a coherent lens on the torrent of data that accompanied the unfolding pandemic.
COMMUNICATION, Health Equity & Policy
- “A Nature survey of more than 7,600 postdoctoral researchers around the world uncovered widespread anxiety and uncertainly about their job paths. Half of the respondents said their satisfaction in their position had worsened in the previous year, and 56% had a negative view on their career outlook. Less than half would recommend a scientific career to their younger self. One-quarter (24%) of respondents said they had experienced discrimination or harassment during their current stint as a postdoc.” In a career feature for Nature, Chris Woolston describes a growing problem for lab science in the UK and EU – a shortage in the pipeline of qualified (and interested) postdoctoral candidates.
- “While other nations have decades of experience in drug price regulation, the IRA creates a new challenge for Medicare’s administrators. How it carries out this work will have important implications for stakeholders and is critical to determining whether this new authority can be sustained or, as some may hope, built upon to regulate additional drug prices.” At Health Affairs Forefront, Ian D. Spatz unpacks the implications of provisions in the recently passed Inflation Reduction Act for drug pricing in the US.
- “Messaging is a core part of public health. From the handwashing reminders in restaurants and hospitals to the safe food handling labels of raw meat, constant reminders of important public health actions are everywhere. This isn’t unique to public health. Even the largest brands know that they need to stay visible or customers lose interest….That’s a lesson that must be heeded today.” In an editorial for STAT News, historian and author Jim Downs and epidemiologist Eleanor J. Murray mine historical precedent to illustrate the importance of clear and well-chosen messaging as a critical component of public health efforts.
- “The most popular AI therapists have millions of users. Yet their explosion in popularity coincides with a stark lack of resources. According to figures from the World Health Organization, there is a global median of 13 mental health workers for every 100,000 people. In high-income countries, the number of mental health workers is more than 40 times higher than in low-income countries. And the mass anxiety and loss triggered by the pandemic has magnified the problem and widened this gap even more.” At Wired, Grace Browne scrutinizes the potential – and the serious limitations – of AI-powered chatbots for mental health applications.