In today’s Duke AI Health Friday Roundup: machine learning deduces physical law; marking a somber COVID milestone; rebuilding trust in public institutions; lack of diversity still a problem for clinical research; frameworks for evaluating clinical AI; digital ID can leave most vulnerable behind; study compares vaping vs. nicotine patches for quitting smoking; European digital privacy protections poised to go beyond GDPR; Great Pacific Garbage Patch turns out to be surprisingly rich in marine life; much more:
- “The patch was less a garbage island than a garbage soup of plastic bottles, fishing nets, tires and toothbrushes. And floating at its surface were blue dragon nudibranchs, Portuguese man-o-wars, and other small surface-dwelling animals, which are collectively known as neuston.” Life finds a way: a story by the New York Times’ Annie Roth reports on the surprise felt by scientists when they realized the a giant gyre of floating oceanic trash – the Great Pacific Garbage Patch – is actually teaming with living creatures.
- “There are figures with human features, a coiled snake with a tail rattle and forked tongue and a 10-foot-long serpent winding its way across the expanse. Some incorporate the features of the ceiling into their design, such as the serpent that appears to emerge from a natural fissure.” An article by the New York Times’ Christine Hauser describes how a team of researchers, equipped with 3-D imaging technology, were able to reveal a trove of ancient Alabama cave art that had previously eluded archaeologists’ eyes.
AI, STATISTICS & DATA SCIENCE
- “…when it comes to orbiting planets, sloshing fluids and dividing cells, concise equations drawing on a handful of operations are bafflingly accurate. It’s a fact that the Nobel laureate Eugene Wigner called “a wonderful gift we neither understand nor deserve” in his 1960 essay “The Unreasonable Effectiveness of Mathematics in the Natural Sciences.” An article in Quanta by Charlie Wood describes what may be the dawning of an era of “GoPro physics” in which machine learning is unleashed on observations in the field or in the lab and derives the underlying principles that govern those phenomena.
- “Through consultation and consensus with a range of stakeholders, we developed a guideline comprising key items that should be reported in early-stage clinical studies of AI-based decision support systems in healthcare.” An article published in Nature Medicine by Vasey and colleagues describes a framework for evaluating clinical AI decision-support systems.
- “This debate is not just academic; the outcome has serious consequences. Today you can be turned down for a job because a so-called emotion-reading system watching you on camera applied artificial intelligence to evaluate your facial movements unfavorably during an interview.” An article in Scientific American by Lisa Feld Barrett explains why AI applications that tout the ability to discern emotional states from facial expressions may be on scientifically shaky ground.
- “The authors use text and image “embedding models” which are used to map texts and images to mathematical vectors that facilitate processing such complex data at scale. For instance, the authors used PDQ hashing to construct a summary mosaic of about 35,000 images submitted to the tipline.” A post at the blog of the University of Michigan’s Artificial Intelligence Laboratory by PhD candidate Ashkan Kazemi describes how a process by which text scooped up from public WhatsApp groups was matched with reports from a misinformation tipline, thereby shedding light on the nature and reach of misinformation on an encrypted app that is otherwise opaque to such scrutiny.
- “Identity verification systems were designed with privileged users in mind; their developers assumed that claimants would own a computer and understand how to use it, that they would be literate, and speak English. Yet for many populations who became unemployed during the pandemic — particularly those that were low-income, elderly, disabled, and non-English speaking — these assumptions created problems from the outset.” An article at Data & Society’s Points blog by Michele Gilman highlights how the use of digital identify verification in state benefit and welfare programs can disadvantage the people who need access to those services the most.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “More Americans have died of Covid-19 than in two decades of car crashes or on battlefields in all of the country’s wars combined. Experts say deaths were all but inevitable from a new virus of such severity and transmissibility. Yet, one million dead is a stunning toll, even for a country the size of the United States, and the true number is almost certainly higher because of undercounting.” A somber New York Times feature illustrates the astronomical toll of COVID on the lives of Americans.
- Findings from a randomized trial of e-cigarettes vs nicotine patches for smoking cessation, reported in Nature Medicine by Hajek and colleagues, show similar outcomes across study arms in terms of effectiveness and safety (with the exception of the incidence of low infant birthweight among pregnant participants, which was lower for those on the e-cigarette arm), although some confounding due to use of e-cigarettes by study participants may have affected the results.
- “A ProPublica investigation into the company’s operations in Nevada, including a review of more than 3,000 pages of internal emails obtained through public records requests, shows the Chicago laboratory’s testing was unreliable from the start….Ultimately, state public health officials found that Northshore’s PCR tests missed 96% of the positive cases from the university campus — errors that sent people infected with COVID-19 back into the community.” ProPublica’s Anjeanette Damon reports on how a COVID testing service that missed virtually all of the truly positive samples tested was able to continuously expand its operations.
- “With rare conditions, it’s often not clear why one child is infected without harm while another battles severe disease. Despite decades of research into polio, it’s not known why the viruses that cause it spare 199 children they infect on average, but cripple one. Scientists have been studying AFM for eight or so years and can’t answer the why question either.” STAT News’ Helen Branswell reports on a recent mysterious outbreak of hepatitis among children and what it may tell us about a different, poorly understood viral illness known as Acute Flaccid Myelitis.
COMMUNICATIONS & DIGITAL SOCIETY
- “…beyond childhood, the brain continues to change and update itself through the modification of its synapses. The evolutionary history of the human brain is written in the genome, but the uniqueness of each mind is written in its unique and ever-changing synaptic circuitry. It seems that what makes us all human is also what makes us all different.” An essay in Quanta by William A. Harris explores the aspects of our brains’ processes that are uniquely human – and uniquely individual and personal as well.
- “Sitkin said courage is needed to counteract the dynamic of people shutting each other out, because reversing distrust involves a willingness to undertake the risk involved in trusting.” An interview with Duke Fuqua School of Business professor Sim Sitkin addresses rebuilding public trust in institutions.
- “’Excited delirium’ is a controversial term that is used by some to describe a person who experiences an acute, extreme disruption in their behavior and ability to think, and often comes up in relation to people who have died in police custody. People with excited delirium are often said to display ‘superhuman strength.’ But most medical authorities do not consider excited delirium to be real.” In his latest ColorCode podcast for STAT News, Nicholas St. Fleur examines the use of a controversial diagnosis known as “excited delirium” and how it has shaped the response of law enforcement in interactions with persons of color.
- “While the European Union’s General Data Protection Regulation is seen as the most comprehensive regulation in the world on the issue of privacy, there’s a slew of other legislation pending, such as the Digital Markets Act and the soon-to-be-voted-on Digital Services Act and plans by the EU to regulate AI, a move that may cost companies up to 6 percent of global revenues. The Markup’s Jeremy Wagstaff reports on European efforts to expand data privacy protections.
- “…the economic cost of long COVID reinforces the value of comprehensive actions to prevent and treat new infections. Mask mandates are unpopular in many areas and a substantial share of the public resists being vaccinated—though each action should still be encouraged. But additional progress might also be made through expanding rapid COVID-19 test capability, global surveillance to detect new SARS-CoV-2 variants, and immediate action should any such variants be detected.” An essay by David M. Cutler in JAMA Health Forum addresses the economic shadows of “long COVID.”
- “…clinical research faces a critical shortcoming. Currently, large swaths of the U.S. population, and those that often face the greatest health challenges, are less able to benefit from these discoveries because they are not adequately represented in clinical research studies. While progress has been made with representation of white women in clinical trials and clinical research, there has been little progress in the last three decades to increase participation of racial and ethnic minority population groups.” A new report from the National Academies of Science, Engineering & Medicine places a spotlight on a persistent problem afflicting clinical research: a lack of diversity among research participants.