AI Health
Friday Roundup
The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.
September 26, 2025
In this week’s Duke AI Health Friday Roundup: promise and peril from medical chatbots; AI and magical thinking; new class of antibiotics shows promise vs MDR strains; delegating tasks to AI may boost dishonesty; dynamics of human-AI teamwork; worries that hospital-acquired infections aren’t being reported; why cats graze; survey reflects public jitters about AI’s prospects; more:
AI, STATISTICS & DATA SCIENCE
- “We demonstrate that collaboration ability is distinct from individual problem-solving ability. Users better able to infer and adapt to others’ perspectives achieve superior collaborative performance with AI—but not when working alone. Moreover, moment-to-moment fluctuations in perspective taking influence AI response quality, highlighting the role of dynamic user factors in collaboration. By introducing a principled framework to analyze data from human-AI collaboration, interactive benchmarks can better complement current single-task benchmarks and crowd-assessment methods.” A research article by Riedl and Weidmann, available as a preprint from PsyArXiv, examines the dynamics of human-AI interactions on shared tasks.
- “Talking to CaBot was both empowering and unnerving. I felt as though I could now receive a second opinion, in any specialty, anytime I wanted. But only with vigilance and medical training could I take full advantage of its abilities—and detect its mistakes. A.I. models can sound like Ph.D.s, even while making grade-school errors in judgment…. The difference between A.I. and earlier diagnostic technologies is like the difference between a power saw and a hacksaw. But a user who’s not careful could cut off a finger.” In a fascinating article for the New Yorker, Druv Khallar examines both the remarkable benefits offered by medical chatbots and the potential perils of using them incautiously.
- “As machine agents become widely accessible to anyone with an internet connection, individuals will be able to delegate a broad range of tasks without specialized access or technical expertise. This shift may fuel a surge in unethical behaviour, not out of malice, but because the moral and practical barriers to unethical delegation are substantially lowered. Our findings point to the urgent need for not only technical guardrails but also a broader management framework that integrates machine design with social and regulatory oversight.” A research article published in Nature by Köbis and colleagues reports findings from several studies that suggest delegating tasks to AI systems can lead to dishonesty.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- Finally, an answer to why cats chow down on grass, only to hork it up later: “…Nicole Hughes, a plant biologist at High Point University, focused not on cats, but on the greens they eat. ‘I know what grass looks like—it’s spiky and likely to snag,’ she says. A mom to two tuxedo cats—Mildred and Merle—Hughes has been saving their hairballs for years, waiting for the right research project to come along.” Science Insider’s Christa Lesté-Lasserre has the story.
- “…we have developed an efficient, high-yielding synthesis for Novltex, a novel class of antibiotics, utilizing low-cost, commercially available building blocks and achieving faster coupling times. This cost-effective methodology enabled the synthesis of a series of Novltex analogues, which were evaluated for their antibacterial activity against MRSA and multidrug-resistant (MDR) bacterial isolates.” A research article published in the Journal of Medicinal Chemistry by Malkawi and colleagues describes a new class of antibiotic compounds that may prove effective against bacterial strains that have developed multidrug resistance.
- “Some hospital leaders, realizing that if they don’t look for infections, they won’t find them, are discouraging testing, clinicians told STAT. Without testing, patients may not get the best treatment. Executives have also pressured staffers in some cases not to report infections to the government, they said….What’s more, regulators have known about the problem for a decade, but neither Medicare nor Congress has taken action to fix it.” An alarming story from STAT News’ Tara Bannow reports on the possibility that some hospitals may be deliberately avoiding testing patients for hospital-acquired infections to avoid being penalized by regulators.
COMMUNICATIONS & Policy
- “The better way forward, I’d wager, is both simple and drastic: to consider A.I. as the constructed piece of code it fundamentally is, not a mystical black box with unlimited potential. There is so much good that A.I. can achieve, from predicting diseases to automating repetitive tasks and translating But if we focus on its supposed mystical aspects…then we not only dupe ourselves but also waste its strengths, locking ourselves once more into the historical cycle of magical thinking.” At the New York Times, scholar of Spiritualism Cody Delistraty scrutinizes the magical thinking accompanying the rise of AI technologies and notes some historical parallels with other technological upheavals.
- “Asked about the effects the broadening use and application of AI systems is likely to have on valued human traits in the coming decade, the public has far more negative than positive views in regard to the future of people’s social intelligence, metacognition, capacity and willingness to think deeply about complex concepts, human agency (the ability to act independently in the world), self-identity and purpose, and their trust in widely shared values and cultural norms.” A report created as part of Elon University’s Imagining the Digital Future project surveys public attitudes and perceptions about the use of artificial intelligence and how it might impact their lives. The report was also featured in a recent research symposium hosted by RTI that also included speakers from Duke University Law School and Duke AI Health, reported on here by WRAL.
- “With the rise of AI coding tools, software engineers have been the subject of a lot of discussion—both in the media and research.…Since late 2022, early-career software engineers (between 22 and 30 years old) have experienced a decline in employment. At the same time, mid-level and senior employment has remained stable or grown. This is happening across the most AI-exposed jobs, and software engineering is a prime example.” An article at IEEE Spectrum news by Gwendolyn Rak reports on recent findings published by the Stanford Digital Economy Lab the effects of AI uptake on entry-level jobs in programming and software engineering.
