Commentary: AI could help to implement health policy
Complex healthcare policies are often challenging to implement. That’s a reality states across the country will deal with as new Medicaid work requirements get rolled out next year.
The new federal rule, part of the Budget Reconciliation Act of 2025 (HR 1), requires adults enrolled under the Affordable Care Act (ACA) expansion of Medicaid eligibility to complete at least 80 hours per month of qualifying work or community engagement activities or meet exemption criteria to maintain their coverage starting Jan. 1.
A new Special Communication published Aug. 7 in JAMA Health Forum suggests that artificial intelligence may enhance implementation of this policy by helping state Medicaid agencies keep eligible individuals enrolled in Medicaid. Going forward, the use of AI can help states and local governments roll out and improve other healthcare policies, too.
“Medicaid work requirements introduce administrative complexities into an already very complex program,” said Dr. Beth McGinty, professor of population health sciences at Weill Cornell and co-founding director of the Cornell Health Policy Center. “It's already very complicated to figure out how to apply for Medicaid and stay enrolled. It differs by state. Forms are confusing, and this new requirement creates additional administrative complexity for Medicaid agencies and the people who are trying to get or stay enrolled.”
Taking the Burden Off Enrollees
The biggest concern is that people who are working or qualify for exemptions to the rule may still lose coverage because they have difficulty submitting documentation. The law requires states to use existing databases to verify eligibility whenever possible, but in many cases states may not be able to verify compliance or exemption status based on the information that’s available to them. In that case, enrollees will need to submit documentation themselves. The Special Communication cited previous research finding that some eligible Arkansas enrollees – when presented with a similar work requirement – lost coverage due to documentation difficulties.
Dr. McGinty and her co-authors Drs. Yongkang Zhang, Fei Wang and William Schpero, as well as Dr. John Ayanian of the University of Michigan, suggest states can use AI tools to take better advantage of information they already have.
“For example, states can use AI to link Medicaid enrollment records with payroll and tax data or enrollment data for other public programs,” said Dr. Schpero, an associate professor in population health sciences at Weill Cornell Medicine and associate center director of the Cornell Health Policy Center. “This would allow agencies to verify if someone is working or qualifies for an exemption without asking the enrollee to fill out more forms or submit more documentation.”
AI tools embedded in online Medicaid application portals could also be used to figure out where enrollees and applicants are having the most difficulty navigating the process, he said.
Today, about a quarter of state Medicaid programs are already using AI chatbots for consumer assistance, said Dr. McGinty, which could be extended to help with implementation of the new work requirements. AI-powered digital assistants could explain requirements, answer questions and help applicants better understand the documentation they need to provide.
Dr. Schpero said AI could also help agencies learn where bottlenecks are in real time by analyzing call center transcripts, help desk messages and activity on Medicaid websites.
“Using AI to provide real-time learning for state Medicaid programs has promise,” Dr. Schpero said. “For example, AI could mine anonymized call center transcripts to understand week to week what challenges enrollees are facing in trying to comply with work requirements, and state could adapt their outreach and assistance programs accordingly.”
Of course, there are some caveats. States and jurisdictions have different levels of AI maturity and IT infrastructure, Dr. McGinty said. The authors argue that federal assistance will be vital for lower-capacity states to implement this approach.
Finally, like with use of other AI tools, there will have to be a human component for it to be successful, Dr. McGinty said.
“There are biases baked into our data that AI implementation will 100% reproduce here, and so having a human in the loop and really careful monitoring and oversight of AI is needed,” she said. “This cannot be a ‘hand it over to the bots’ solution.”
Journal
JAMA Health Forum
Article Publication Date
7-Aug-2026
'A²SG' developed to enhance spiking neural network performance, achieving world-class accuracy
National Research Council of Science & Technology
image:
(From left) Senior Researcher Seongsik Park (corresponding author), Master's student researcher Kang Yechan (first author), and Master's student researcher Park Sohee (co-author) pose for a commemorative photo at the ICML 2026 poster session held at COEX in Seoul.
view moreCredit: Korea Institute of Science and Technology(KIST)
While asking questions to ChatGPT and similar tools to generate images has become part of daily life, behind this convenience lies the massive power consumption of huge data centers. As power issues have emerged as a societal challenge, global big tech companies are moving to develop low-power "neuromorphic semiconductors" that mimic the human brain, and South Korea has also designated "AI semiconductors" as a priority technology within its national strategic technology framework to secure self-reliance. The human brain operates on minimal energy by exchanging signals only when necessary. "Spiking neural networks (SNNs)," modeled after the human brain, consume far less power than conventional AI, but they have struggled to match the performance of deep neural networks (DNNs)-the foundation of current AI systems like ChatGPT.
A research team led by Senior Researcher Seongsik Park of the Semiconductor Technology Research Division at the Korea Institute of Science and Technology (KIST; President Oh Sang-rok) announced that it has developed "A²SG," a new learning technique that enhances the learning performance of SNNs. This research was accepted as a regular paper at ICML 2026, one of the world's top three artificial intelligence conferences, and was presented on Tuesday, July 7, at the conference (July 6-11, COEX, Seoul)-the first time in the conference's history that it was held in South Korea.
The A²SG developed by the research team was applied to a large-scale spiking neural network (SNN) based on the transformer-the core architecture of ChatGPT-and achieved world-leading accuracy for spiking neural networks in the large-scale ImageNet image recognition evaluation. Artificial intelligence refines its models by solving problems, verifying results, and iteratively finding better solutions. In this process, "gradients" serve to indicate the direction and extent to which each model should be adjusted. However, because SNNs exchange signals differently from DNNs, it was difficult to fine-tune the model adequately using only the gradients employed in conventional DNNs. A²SG enhanced training performance by combining an "adaptive" approach-which adjusts the tuning method based on the training context-with an "asymmetric" approach that reflects the characteristics of brain neurons.
A²SG is also noteworthy for delivering world-class performance at a lower cost. It achieved higher accuracy while requiring only about one-sixth the computational overhead of Google's leading training method, and demonstrated consistent performance improvements across a wide range of neural network architectures and applications-from small to large models-thus proving its versatility.
This achievement is significant in that KIST has secured core technology in the field of neuromorphic AI learning algorithms, an area previously led by universities and global tech giants. Since A²SG can be implemented using software alone without any hardware modifications, it can be applied to low-power applications such as on-device AI in smartphones, wearable devices, and drones, as well as smart sensors that operate 24 hours a day. Going forward, the research team plans to commercialize low-power AI semiconductors following large-scale model training and verification of neuromorphic hardware. They also intend to apply the acquired training algorithm technology to the development of AI models for next-generation AI semiconductors, such as the probability-based "RPU (Random Processing Unit)" currently under development at KIST.
Seongsik Park , a senior researcher at KIST, stated, "This research addresses the structural challenges in learning that had hindered the performance improvement of neuromorphic AI, thereby increasing the potential for the practical application of low-power AI." He added, "Going forward, we will develop this core technology into AI models that operate on next-generation AI semiconductors, contributing to the dawn of the low-power AI era." The KIST Post-Silicon Semiconductor Institute, which has made it its mission to develop innovative semiconductors that will transform future computing paradigms, plans to link this achievement with its research on next-generation AI semiconductors to secure core foundational technologies for high-efficiency intelligent semiconductors and strengthen the foundation for technological self-reliance.
###
KIST was established in 1966 as the first government-funded research institute in Korea. KIST now strives to solve national and social challenges and secure growth engines through leading and innovative research. For more information, please visit KIST’s website at https://kist.re.kr/eng/index.do
This research was supported by the Ministry of Science and ICT (Minister Bae Kyung-hoon) of Korea through the KIST Institutional Program (26E0020), the Institute of Information & Communications Technology Planning & Evaluation (IITP) (RS-2025-02218733), and the Sejong Science Fellowship of the National Research Foundation of Korea (NRF-2021R1C1C2010454). The findings of this study are scheduled to be published in the latest issue of the Proceedings of the International Machine Learning Conference (PMLR).
Article Title
A²SG: Adaptive and Asymmetric Surrogate Gradients for Training Deep Spiking Neural Networks
Dr. Seongsik Park 's research team at KIST is explaining the details of the A2SG study to researchers at the ICML poster session held at COEX in Seoul.
Conventional artificial intelligence (DNN) has a gentle learning landscape (left). The middle and right figures show the same learning landscape for the same spiking neural network (SNN), where the search path varies depending on the compass (learning method) used. While the conventional method reaches a steep and rugged point (middle), A²SG reaches a flat point, resulting in improved performance (right). Each figure depicts the terrain surrounding the point reached during learning.
(a) Valid intervals of the surrogate gradient and the asymmetric (ASY) function form (b) How the spatiotemporal adaptive surrogate gradient operates by detecting changes in the learning landscape in real time and self-correcting accordingly (c) A schematic diagram showing the vicinity of the point reached by the training process before and after applying A²SG on the loss landscape
This image (t-SNE) displays the features extracted from images by artificial intelligence on a two-dimensional map, where each point represents a single image and the color indicates the type of object. The model trained using the conventional method (left) shows categories mixed together, making them difficult to distinguish, whereas the model trained with A²SG (right) clearly separates the categories. This demonstrates that A²SG enables stable and accurate training of SNN models.
Credit
Korea Institute of Science and Technology(KIST)
JMIR News: Drones, misinformation regulation, and the dispute over specialized AI models
JMIR Publications
(Toronto, August 7, 2026) — JMIR Publications released three feature News and Perspectives articles on current issues and developments in digital health care.
OpenEvidence and UpToDate Challenge Comparison Study
In “Generalist Versus Specialist: What Is the Best AI Model for Health Care?”, JMIR Correspondent Simon Spichak investigates the controversy surrounding recent research comparing the clinical performance of specialized medical LLMs with generalist frontier models. NYU Langone researchers found that frontier models outperformed the OpenEvidence and UpToDate models in answering benchmark questions, prompting critique from both companies on the study authors’ alleged undisclosed conflicts of interest and methodological issues. Beyond the question of the study’s validity, however, Spichak raises the issue of whether using AI tools for clinical decision support makes a positive difference. “Despite studies showing how well these products work on benchmarks,” he writes, “there is a lack of data linking them to real-world clinical outcomes.”
Please cite as:
Spichak S. Generalist Versus Specialist: What Is the Best AI Model for Health Care?
J Med Internet Res 2026;28:e108066
URL: https://www.jmir.org/2026/1/e108066
DOI: 10.2196/108066
China’s Crackdown on Online Misinformation
Misinformation regulation has entered a new chapter, says consumer tech consultant and analyst Tim Bajarin in his op-ed "China Has Moved to Regulate Expertise Online—and the West Should Pay Attention". China’s regulations now require platforms to verify users claiming expertise in certain professional areas by providing proof of their credentials, or otherwise risk fines and restrictions, “treating this kind of online speech as something that can be licensed, audited, and penalized.” Though Bajarin notes that this degree of state-run moderation of online content conflicts with democratic ideals, he argues that the status quo elsewhere in the world has so far failed to address some of the root causes driving online misinformation. “The United States treats misinformation primarily as a content problem,” he writes. “China treats it more as a systems problem.”
Please cite as:
Bajarin T. China Has Moved to Regulate Expertise Online—and the West Should Pay Attention.
J Med Internet Res 2026;28:e107872
URL: https://www.jmir.org/2026/1/e107872
DOI: 10.2196/107872
Drones to Deliver Prescription Medications in Underserviced Areas
For many residents of areas with few pharmacies, picking up prescriptions can be costly, difficult, and time-consuming—but this may soon change. JMIR Correspondent Jenna Congdon reports on recent and upcoming drone prescription delivery initiatives in "Drones in the (Pharmacy) Desert: Can Prescription Delivery via Drone Improve Health Care Access?". She speaks with Bri Brown Robinson on the Cleveland Clinic’s drone-based pharmacy delivery program, discussing the complexities of aviation regulations, patient privacy, environmental variables, and more. “From a clinical workflow perspective,” writes Congdon, “integrating drone delivery into existing pharmacy systems requires robust solutions for prescription verification, chain of custody management, and patient-to-clinician communication.” Despite these logistical challenges, drone delivery has the potential to improve health care access for individuals in remote or underserviced locations.
Please cite as:
Congdon J. Drones in the (Pharmacy) Desert: Can Prescription Delivery via Drone Improve Health Care Access?
J Med Internet Res 2026;28:e108067
URL: https://www.jmir.org/2026/1/e108067
DOI: 10.2196/108067
About JMIR Publications News and Perspectives
JMIR Publications is a leading open access publisher of digital health research. The News and Perspectives section is the newest addition to its portfolio, established to bring the rigor and integrity of academic publishing to scientific journalism. The section features well-researched, expert-driven content from the Scientific News Editor, Kayleigh-Ann Clegg, PhD, and a network of specialist JMIR Publications Correspondents to keep the digital health community informed, inspired, and ahead of the curve.
About JMIR Publications
JMIR Publications is a leading open access publisher of digital health research and a champion of open science. With a focus on author advocacy and research amplification, JMIR Publications partners with researchers to advance their careers and maximize the impact of their work. As a technology organization with publishing at its core, we provide innovative tools and resources that go beyond traditional publishing, supporting researchers at every step of the dissemination process. Our portfolio features a range of peer-reviewed journals, including the renowned Journal of Medical Internet Research.
To find out more about JMIR Publications, visit jmirpublications.com or connect with them on Bluesky, X, LinkedIn, YouTube, Facebook, and Instagram.
Media Contact:
Dennis O’Brien, Vice President, Communications & Partnerships
JMIR Publications
communications@jmir.org
+1 416-583-2040
The content of this communication is licensed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, published by JMIR Publications, is properly cited.
Journal
Journal of Medical Internet Research
Method of Research
Commentary/editorial
Subject of Research
People
Article Publication Date
31-Jul-2026
No comments:
Post a Comment