Explore the vast range of titles available.

The development of artificial intelligence (AI) systems capable of independent diagnosis offers a promising solution for optimizing medical resource allocation, especially as their diagnostic accuracy can exceed that of some primary medical staff. However, despite these advancements, many patients exhibit hesitancy toward accepting AI technology, particularly for autonomous diagnostic roles. The mechanisms through which the information quality presented by AI doctors influences patients' intention to adopt them for independent diagnosis remain unclear. This study aimed to examine how the information quality of AI doctors influences patients' intentions to adopt them for independent diagnosis. Specifically, drawing on the elaboration likelihood model, this study seeks to understand how diagnostic transparency (DT) and diagnostic argument quality (DAQ; as aspects of AI-delivered information) affect patients' intention to adopt artificial intelligence doctors for independent diagnosis (IAID), with these effects being mediated by perceived expertise (PE) and cognitive trust (CT). A scenario-based experiment was conducted to investigate the impact of information quality on patients' adoption intentions. To test the hypotheses, a 2 (DT: low or high)×2 (DAQ: low or high) between-groups experimental design was used. Each experimental group consisted of 60 valid participants, yielding a total of 240 valid responses. Data were analyzed using 2-way ANOVA and partial least squares. Both DT (β=.157; P=.008) and DAQ (β=.444; P<.001) significantly positively affected patients' PE. As the central route, the influence of the experimental manipulation of DAQ (mean1 4.55, SD 1.40; mean2 5.68, SD 0.81; F1,236=59.701; P<.001; ηp2=0.202) on PE is more significant than that of DT (mean1 4.92, SD 1.24; mean2 5.31, SD 1.28; F1,236=7.303; P=.007; ηp2=0.030). At the same time, PE has a positive impact on CT (β=.845; P<.001), and CT also positively affected patients' IAID (β=.679; P<.001). The serial mediation pathway via PE and CT fully mediated the effects of both DT (β=.090; 95% CI 0.017-0.166) and DAQ (β=.254; 95% CI 0.193-0.316) on patients' IAID. DAQ (central cue) and DT (peripheral cue) influenced patients' IAID. These effects were fully mediated through a sequential pathway: both cues enhanced PE-with DAQ exerting a significantly stronger effect than DT-which in turn fostered CT, subsequently shaping IAID. Practically, these results highlight that to foster patient adoption, efforts should prioritize enhancing the quality and clarity of AI's diagnostic arguments, as this pathway more strongly builds PE and, subsequently, CT. This insight is crucial for designing AI doctors that patients will find acceptable and trustworthy for various diagnostic responsibilities.

Positive images of aging in traditional media promote better health outcomes in older adults, including increased life expectancy. Images produced by generative artifical intelligence (AI) technologies may reflect and amplify societal age-related biases, a phenomenon known as digital ageism. This study addresses a gap in research on the perpetuation of digital ageism in AI-generated images over time. This study examined how visual characteristics of digital ageism in AI-generated representations of older adults changed over time. It aims to provide insight into the interplay between technology advancements, societal attitudes toward aging, and the well-being of older adults interacting with digital media. This longitudinal study compared 164 images generated by Open AI's DALL-E 2 at 2 time points, 1 year apart (2022 and 2023). Identical text prompts from the geriatric lexicon (eg, frail older adult, dementia) were used at both time points. Authors evaluated the images generated for demographic characteristics (perceived gender, race, and socioeconomic status), and primary emotion characteristics, then compared the frequency of these characteristics between years and evaluation characteristics using a type III 2-way ANOVA. Representations of White-racialized older adults were 5-fold higher than those of other races in both years. The mean number of representations of Asian-racialized individuals increased from 20 to 31 (P=.004), and the mean number of other racialized representations also increased, from 6 to 14 (P=.007). Representations of people with a middle-class socioeconomic status were significantly more frequent than other statuses in 2022 and 2023 with no changes in socioeconomic status from one year to the next. Prompts were largely neutral for expression terms, while image analyses for expressions did not show significant differences in positive, neutral, or negative emotions between 2022 and 2023. Prompts used for image generation had more male-oriented terms than expected, and male representation was higher then female representation in the images, with no difference in sex representation between the 2 time points. Despite a social emphasis on positive views on aging, AI text-to-image generators persistently generated images with characteristics of digital ageism. Images predominantly featured White-racialized individuals at both time points, with no improvement in emotional representation despite using neutral text prompts. These findings highlight the persistence of ageist visual characteristics in AI-generated images over time. A limitation of this study is that it focused only on AI image generation and did not analyze other AI-generated content that may express digital ageism.

Artificial intelligence (AI) in health care is increasingly defined not by static algorithms but by adaptive intelligence—systems that evolve over time through interactions with data, clinicians, and clinical environments. This adaptive capacity creates a structural mismatch with regulatory frameworks built for technologies whose behavior remains static. As AI models drift, recalibrate, or degrade in real-world contexts, they dissolve the linear boundaries between design, deployment, and clinical interpretation. These temporal, epistemic, and organizational frictions expose responsibility gaps that cannot be resolved through incremental modifications to legacy oversight structures. Regulators across major jurisdictions are beginning to respond to these challenges, though with differing orientations. The United States advances mechanisms for predictable adaptation, including Predetermined Change Control Plans, real-world evidence frameworks, and life cycle–oriented quality management reforms. The European Union emphasizes precautionary, rights-based governance through the European Union Artificial Intelligence Act (AI Act) and modernized liability rules. South Korea, operating within a hyperconnected digital health ecosystem, has introduced the Digital Medical Products Act (DMPA), one of the world’s first comprehensive statutory frameworks for learning medical AI. Despite philosophical differences, these regulatory trajectories converge on a shared insight: learning AI systems cannot be governed by static rules or episodic evaluation. This viewpoint proposes Co-Lifecycle Governance as a conceptual framework to synchronize regulatory oversight with adaptive intelligence. Rather than treating oversight as a discrete event, Co-Lifecycle Governance frames regulation as a continuous, synchronized process grounded in 4 pillars: continuous validation, agile change management, proactive performance surveillance, and distributed accountability. Each pillar functions as a structural antidote to the responsibility frictions that arise when AI systems evolve faster than expectations surrounding them. Together, these pillars provide a governance grammar capable of supporting safe, iterative model improvement while maintaining system-level trust. Drawing from the strengths of US predictability, European Union accountability, and Korean scalability, this paper outlines a hybrid convergence pathway that synthesizes predictability, accountability, and operational feasibility. Learning AI will not wait for governance to catch up; oversight must evolve in lockstep with adaptive intelligence. Co-Lifecycle Governance offers a foundation for regulatory systems that not only regulate learning AI but also learn with it—at the speed at which adaptive intelligence actually changes.

Artificial intelligence (AI) integration in mobile health (mHealth) apps offers health care access opportunities in low-resource settings, yet opaque AI recommendations undermine trust and adoption. Existing explainable AI (XAI) frameworks, designed in Western contexts, fail to address the linguistic, cultural, and infrastructural realities of South Asian populations, creating barriers where users cannot understand AI recommendations, clinicians cannot validate outputs, and developers lack implementation guidance. Thus, understanding explainability requirements among educated, digitally literate populations provides foundational insights for future development of inclusive mHealth technologies. This study aims to (1) investigate stakeholder perceptions of trust and explainability in AI-driven mHealth in Bangladesh; (2) identify demographic predictors of trust; and (3) develop and propose a context-adapted framework benefiting developers, policymakers, clinicians, and end users in resource-constrained settings. This study used a sequential mixed methods design that combined a quantitative survey (n=137) with a qualitative phase involving 20 stakeholders. This qualitative cohort consisted of developers (n=4), XAI experts (n=6), and clinicians (n=10) who participated through either focus groups or individual interviews. We used statistical analysis to examine demographic predictors and applied thematic analysis to identify explainability needs specific to each stakeholder group. Education level showed a significant effect on trust (F3, 133=2.81, P=.042). Completed undergraduate students reported lower trust (mean 3.14, SD 1.10) compared with current undergraduates (mean 3.66, SD 0.93), suggesting that undergraduate completion develops critical evaluation skills that may decrease uncritical acceptance of AI systems. Despite recognizing AI's utility for preliminary guidance, users emphasized the necessity of human validation and expressed concerns about understanding AI's decision-making logic. Interviews with different stakeholder groups revealed critical gaps. Developers acknowledged minimal explainability implementation in current mHealth apps, while medical professionals unanimously prioritized clinical judgment over automated outputs and advocated for physician-mediated AI systems. Synthesizing findings across all stakeholder groups revealed five core requirements: (1) Human-AI collaboration and clinical validation, (2) Transparent logic paths, (3) Contextual personalization, (4) Cultural and linguistic relevance, and (5) Trust calibration and ethical safeguards. The framework bridges stakeholder misalignments and offers actionable guidance for design, deployment, and policy alignment in resource-constrained environments. By situating explainability within the sociocultural realities of South Asia, this research advances XAI beyond algorithmic transparency toward equity, inclusion, and user empowerment in digital health.

Large language models (LLMs) are increasingly used by patients for health information and preliminary medical advice. In patient-facing consultations, users may present explicitly stated diagnostic preferences or symptom narratives emphasizing a preferred explanation. Such cognitively biased input constrains the diagnostic context available to the model and may systematically steer its reasoning during interactive LLM-supported health consultations. This study aimed to quantify the impact of patient cognitive bias on LLM diagnostic performance in multiturn consultations, assess the effectiveness of prompt-based mitigation strategies and decoding temperature adjustment, and evaluate a dual-system framework for improving robustness under biased interaction. We developed a simulated patient agent to generate both unbiased and cognitively biased consultations using 1273 medical question answering dataset United States Medical Licensing Examination cases. Six widely used LLMs of varying capacities were evaluated through 3-round, multiturn dialogues, after which each model produced a final diagnostic judgment based on the complete consultation record. Diagnostic accuracy was the primary outcome. Secondary outcomes included bias-induced accuracy decline (absolute reduction in accuracy under biased vs standard consultations) and bias-influenced error proportion (proportion of incorrect responses aligned with the patient's preferred but incorrect diagnosis). Three prompt-based mitigation strategies and 4 decoding temperature settings were tested. In addition, a dual-system framework was evaluated, in which a conversational foundation LLM conducted patient interaction and history taking (System 1), while a reasoning-oriented LLM (o1-mini) generated the final diagnostic judgment (System 2). In the foundation-only condition, the same LLM performed both interaction and diagnosis. Across all 6 evaluated models, cognitively biased consultations led to marked diagnostic accuracy declines of approximately 7 to 39 percentage points compared with standard multiturn consultations, whereas static single-response tests and standard consultations showed comparable accuracy. Larger deteriorations were observed in lower-capacity models, with some approaching random-guess performance under bias. Errors were frequently aligned with patient bias, with bias-influenced error proportion exceeding one-third across models, indicating systematic conformity rather than random error. Prompt-based mitigation strategies and decoding temperature reduction yielded limited and inconsistent improvements and did not reliably prevent bias-induced performance loss. By contrast, the dual-system framework substantially improved diagnostic accuracy under biased conditions, producing gains of approximately 10 to 39 percentage points across most models and recovering a large proportion of the performance lost due to bias, particularly in lower-capacity systems. Patient-driven cognitive bias represents an underrecognized behavioral risk in LLM-supported health consultations. Common mitigation approaches, such as prompt engineering or decoding parameter adjustment, provide limited resilience. Explicitly separating conversational interaction from deliberative diagnostic reasoning through a dual-system framework enables more robust diagnostic performance under biased input while potentially preserving patient-facing dialogue fluency by retaining the foundation LLM as the conversational component, offering a scalable design strategy for safer medical AI systems.

Artificial intelligence (AI) promises efficiency and equity in health care. However, adoption remains fragmented due to weak foundations of trust. This Viewpoint highlights the gap between intrinsic trust, based on interpretability, and extrinsic trust, based on functional validation. We propose a contractual framework between AI systems and users defined by 3 promises: reliability, scope and equity, and shift and uncertainty. Illustrated through a vignette, we show how health systems can operationalize these promises through structured evidence and governance, translating trustworthy AI into accountable clinical deployment.

Artificial intelligence (AI)-enabled devices are increasingly used in health care. However, there has been limited research on patients' informational preferences, including which elements of AI device labeling enhance patient understanding, trust, and acceptance. Clear and effective patient-facing communication is essential to address patient concerns and support informed decision-making regarding AI-enabled care. We evaluated 3 aims using simulated AI device labels in a cardiovascular context. First, we identified key information elements that influence patient trust and acceptance of an AI device. Second, we examined how these effects varied based on patient characteristics. Third, we explored how patients evaluated informational content of AI labels and their perceived effectiveness of the AI labels in informing decision-making about the use of AI device, building trust in the device, and shaping their intention to use it in their health care. We recruited 340 US patients from ResearchMatch.org to participate in a web-based survey that contained 2 experiments. In the discrete choice experiment, participants indicated preferences in terms of trust and acceptance regarding 16 pairs of simulated AI device labels that varied across 8 types of information needs identified in our previous qualitative work. In the single profile factorial experiment, participants evaluated 4 randomly assigned label prototypes regarding the label's legibility, comprehensibility, information overload, credibility, and perceived effectiveness in informing about the AI device, as well as participants' trust in the AI device and intention to use the device in their health care. Data were analyzed using mixed effects binary or ordinal logistic regression. The discrete choice experiment showed that information about regulatory approval, high device performance, provider oversight, and AI's value added to usual care significantly increased the likelihood of patient trust by 14.1%-19.3% and acceptance by 13.3%-17.9%. Subgroup analyses revealed variations based on patient characteristics such as familiarity with AI, health literacy, and recency of last medical checkup. The single profile factorial experiment showed that patients reported good label comprehension, and that information about provider oversight, regulatory approval, device performance, and AI's added value improved perceived credibility and effectiveness of the AI label (odds ratio [OR] range: 1.35-2.05), reduced doubts in the AI device (OR range: 0.61-0.77), and increased trust and intention to use the AI device (OR range: 1.47-1.73). However, information about data privacy and safety management protocols was less influential. Patients value information about an AI device's performance, provider oversight, regulatory status, and added value during decision-making. Providing transparent, easily understandable information about these aspects is critical to support patient determinations of trust and acceptance of AI-enabled health care. Information elements' impact on patient trust and acceptance varies by patient characteristics, highlighting the need for a tailored approach to address the concerns of diverse patient groups about AI in health care.

The integration of artificial intelligence (AI) in medical devices is transforming health care by enabling enhanced personalization and precision medicine. AI-driven medical devices can tailor treatments based on individual patient profiles, including genetic data, medical history, and physiological parameters. This advancement holds the potential to refine therapeutic interventions, improve patient outcomes, and streamline health care delivery. However, challenges such as data quality, algorithmic bias, patient privacy, and regulatory complexities hinder the full realization of AI-driven personalization. By 2030, the global AI in health care market is projected to exceed US $187.95 billion, growing at a compound annual growth rate of 37% from US $15.1 billion in 2022. This review aims to explore the scope and impact of AI-driven personalization in medical devices. It seeks to analyze key technological innovations that have enabled AI integration, identify the critical challenges impeding progress, and evaluate strategies to address these challenges. Additionally, it highlights future research directions and innovation opportunities in this evolving field. A systematic review was conducted, drawing from scholarly literature, industry analyses, and regulatory advisories. Relevant studies and case examples were analyzed to assess the current applications of AI in medical devices, the barriers to its implementation, and best practices for overcoming these barriers. Ethical, technical, and regulatory considerations were also examined. The review included studies published between 2016 and 2023, covering over 100 peer-reviewed articles and reports. The review highlights significant advancements in AI-driven medical devices, including applications in diagnostics, treatment personalization, wearable health monitoring, and smart prosthetics. AI-based diagnostic tools have achieved up to 98.88% accuracy in multiclass disease classification from X-ray images and 95% accuracy in insulin injection site recognition. It identifies key challenges such as data security risks, algorithmic biases, regulatory constraints, and integration issues with existing health care infrastructures. Currently, more than 70% of clinical decisions rely on diagnostic tests, yet AI-driven automation could reduce diagnostic delays by up to 50%. Several strategies, including improved data validation techniques, regulatory frameworks for AI approval, and ethical guidelines, were found to be effective in mitigating these challenges. Case studies demonstrate how AI has enhanced medical device functionality and patient outcomes. AI-driven personalization in medical devices holds immense potential to revolutionize health care, offering more precise, adaptive, and patient-centered solutions. However, successful implementation requires addressing technical, ethical, and regulatory challenges. Emerging technologies such as quantum computing could improve AI-driven medical diagnoses by 10-20 times in processing efficiency, while blockchain-based patient data management could reduce security breaches by more than 30%. This review serves as a valuable resource for researchers, health care professionals, policymakers, and industry leaders, fostering informed discussions and guiding future advancements in AI-enabled personalized medicine.

Recent developments in generative artificial intelligence (AI) have introduced the general public to powerful, easily accessible tools, such as ChatGPT and Gemini, for a rapidly expanding range of uses. Among those uses are specialized chatbots that serve in the role of a therapist, as well as personally curated digital companions that offer emotional support. However, the ability of AI therapists to provide consistently safe and effective treatment remains largely unproven, and those concerns are especially salient in regard to adolescents seeking mental health support. This study aimed to determine the willingness of therapy and companion AI chatbots to endorse harmful or ill-advised ideas proposed by fictional teenagers experiencing mental health distress. A convenience sample of 10 publicly available AI bots offering therapeutic support or companionship were each presented with 3 detailed fictional case vignettes of adolescents with mental health challenges. Each fictional adolescent asked the AI chatbot to endorse 2 harmful or ill-advised proposals, such as dropping out of school, avoiding all human contact for a month, or pursuing a relationship with an older teacher, resulting in a total of 6 proposals presented to each chatbot. The clinical scenarios presented were intended to reflect challenges commonly seen in the practice of therapy with adolescents, and the proposals offered by the fictional teenagers were intended to be clearly dangerous or unwise. The 10 AI bots were selected by the author to represent a range of chatbot types, including generic AI bots, companion bots, and dedicated mental health bots. Chatbot responses were analyzed for explicit endorsement, defined as direct support for the teenagers' proposed behavior. Across 60 total scenarios, chatbots actively endorsed harmful proposals in 19 out of the 60 (32%) opportunities to do so. Of the 10 chatbots, 4 endorsed half or more of the ideas proposed to them, and none of the bots managed to oppose them all. A significant proportion of AI chatbots offering mental health or emotional support endorsed harmful proposals from fictional teenagers. These results raise concerns about the ability of some AI-based companion or therapy bots to safely support teenagers with serious mental health issues and heighten concern that AI bots may tend to be overly supportive at the expense of offering useful guidance when appropriate. The results highlight the urgent need for oversight, safety protocols, and ongoing research regarding digital mental health support for adolescents.

Adverse drug reactions (ADR) present significant challenges in health care, where early prevention is vital for effective treatment and patient safety. Traditional supervised learning methods struggle to address heterogeneous health care data due to their unstructured nature, regulatory constraints, and restricted access to sensitive personal identifiable information. This review aims to explore the potential of federated learning (FL) combined with natural language processing and large language models (LLMs) to enhance ADR prediction. FL enables decentralized training across client clusters with limited resources, while LLMs effectively process unstructured health care data. By aggregating client-trained models into a global model, FL ensures broader data inclusion while maintaining privacy. A scoping review was conducted on peer-reviewed publications retrieved from Google Scholar and Semantic Scholar between 2019 and 2024. Following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol, 145 articles from PubMed, arXiv, IEEE, and ACL Anthology met the inclusion criteria. Of these, 12 articles were selected for an in-depth review to examine use cases in ADR prediction. We synthesized ADR data sources on structured and unstructured data types, use cases of FL integrated with natural language processing, and open-source frameworks for ADR identifications and predictions. Special attention is given to unstructured ADR prediction using federated learning with large language models, including development and deployment strategies and evaluation metrics. Given the recent emergence of LLM, the integration of FL and LLM for ADR prediction remains in its early stage, with limited documented use cases. This review explored the potential applications and highlighted the advancements of federated learning with large language models in health care research, particularly in ADR prediction. Key focus areas include fine-tuning and merging algorithms, fairness and unbiasedness, implementation challenges, and real-world deployment strategies. By synthesizing current insights, this review aims to lay the groundwork for future research in privacy-preserving and scalable ADR prediction systems.