Explore the vast range of titles available.

Artificial intelligence-based clinical decision support systems (AI-CDSSs) have enhanced personalized medicine and improved the efficiency of health care workers. Despite these opportunities, trust in these tools remains a critical factor for their successful integration into practice. Existing research lacks synthesized insights and actionable recommendations to guide the development of AI-CDSSs that foster trust among health care workers. This systematic review aims to identify and synthesize key factors that influence health care workers' trust in AI-CDSSs and to provide actionable recommendations for enhancing their trust in these systems. We conducted a systematic review of published studies from January 2020 to November 2024, retrieved from PubMed, Scopus, and Google Scholar. Inclusion criteria focused on studies that examined health care workers' perceptions, experiences, and trust in AI-CDSSs. Studies in non-English languages and those unrelated to health care settings were excluded. Two independent reviewers followed the Cochrane Collaboration Handbook and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines. Analysis was conducted using a developed data charter. The Critical Appraisal Skills Programme tool was applied to assess the quality of the included studies and to evaluate the risk of bias, ensuring a rigorous and systematic review process. A total of 27 studies met the inclusion criteria, involving diverse health care workers, predominantly in hospitalized settings. Qualitative methods were the most common (n=16, 59%), with sample sizes ranging from small focus groups to cohorts of over 1000 participants. Eight key themes emerged as pivotal in improving health care workers' trust in AI-CDSSs: (1) System Transparency, emphasizing the need for clear and interpretable AI; (2) Training and Familiarity, highlighting the importance of knowledge sharing and user education; (3) System Usability, focusing on effective integration into clinical workflows; (4) Clinical Reliability, addressing the consistency and accuracy of system performance; (5) Credibility and Validation, referring to how well the system performs across diverse clinical contexts; (6) Ethical Consideration, examining medicolegal liability, fairness, and adherence to ethical standards;(7) Human Centric Design, pioritizing patient centered approaches; (8) Customization and Control, highlighting the need to tailor tools to specific clinical needs while preserving health care providers' decision-making autonomy. Barriers to trust included algorithmic opacity, insufficient training, and ethical challenges, while enabling factors for health care workers' trust in AI-CDSS tools were transparency, usability, and clinical reliability. The findings highlight the need for explainable AI models, comprehensive training, stakeholder involvement, and human-centered design to foster health care workers' trust in AI-CDSSs. Although the heterogeneity of study designs and lack of specific data limit further analysis, this review bridges existing gaps by identifying key themes that support trust in AI-CDSSs. It also recommends that future research include diverse demographics, cross-cultural perspectives, and contextual differences in trust across various health care professions.

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) 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.

Prior research on human-like virtual profiles (VPs)-computer-generated imagery (CGI; displaying legible artificiality) or artificial intelligence (AI)-generated personas on social media-is often based on small samples and limited data. Furthermore, prior studies group highly photorealistic virtual influencers with abstract virtual characters, complicating comparisons. Recent generative AI has enabled large-scale production of synthetic content. These advances have drastically increased the quality and volume of synthetic media. Much of the prior literature predates these developments, leaving open questions about how posting behavior and VP design now relate to reach and engagement outcomes. This study aimed to examine how posting behavior and key VP attributes, namely, photorealism; physical, behavioral, and narrative consistency; and human copresence related to reach and engagement across content formats (static images vs videos) on Instagram. A total of 157 human-like female VPs were included in the final Instagram dataset. During the initial screening stage, 5 human-like male macro and mega VPs were identified but excluded because their number was too small to support meaningful subgroup comparison. Engagement was operationalized as like rate (likes or followers) for images and videos; reach was measured via examined absolute impressions. Performance was summarized using the single best-performing post and the arithmetic mean of the top 3 posts per format. Profiles and posts were further coded using predefined content variables covering visual realism, identity consistency, copresence structure, appearance patterns, and body-type representation. Higher-performing content was more commonly observed among larger profiles, particularly for video metrics. Among videos exceeding 20 million impressions, out of 12 videos, 7 (58%) were produced by CGI-like VPs, whereas high video like rates were more often observed among photorealistic or face-swapped profiles. In the image engagement analysis, out of 8 highest-like-rate posts, 6 (75%) featured VPs with dark hair and dark eyes, and out of the 8 posts, 4 (50%) included copresence of multiple subjects. Furthermore, of the 157 top-performing profiles reviewed, 122 (77.71%) demonstrated stable visual identity and consistent behavioral and narrative presentation, whereas noticeable inconsistencies were rarely observed in this subset. These findings represent descriptive patterns within established macro- and mega-level profiles rather than as causal predictors of growth. Among established human-like female VPs on Instagram, higher engagement and reach frequently co-occurred with larger profile scale, identity coherence, and transparent virtual presentation than with photorealism alone. CGI-like aesthetics aligned more with algorithmic distribution, whereas photorealistic motion content featured more prominently among posts with high active engagement. These results suggest that audiences may respond more favorably to coherent and legible virtual identities than to ambiguous realism. Because the study was restricted to macro- and mega-level profiles, the observed traits should be interpreted as characteristics of successful incumbents rather than as general predictors of VP growth.

Maris et al document important ethical challenges at the intersection of electronic health record data and artificial intelligence development, but existing governance frameworks designed for secondary data use are categorically insufficient for artificial intelligence model training, which creates persistent deployable artifacts that encode local clinical patterns as generalizable knowledge. Drawing on two decades of stewardship framework development, we propose extending governance from data stewardship to model stewardship.

Technological advancements have significantly reshaped health care, introducing digital solutions that enhance diagnostics and patient care. Artificial intelligence (AI) stands out, offering unprecedented capabilities in data analysis, diagnostic support, and personalized medicine. However, effectively integrating AI into health care necessitates specialized competencies among professionals, an area still in its infancy in terms of comprehensive literature and formalized training programs. This systematic review aims to consolidate the essential skills and knowledge health care professionals need to integrate AI into their clinical practice effectively, according to the published literature. We conducted a systematic review, across databases PubMed, Scopus, and Web of Science, of peer-reviewed literature that directly explored the required skills for health care professionals to integrate AI into their practice, published in English or Spanish from 2018 onward. Studies that did not refer to specific skills or training in digital health were not included, discarding those that did not directly contribute to understanding the competencies necessary to integrate AI into health care practice. Bias in the examined works was evaluated following Cochrane's domain-based recommendations. The initial database search yielded a total of 2457 articles. After deleting duplicates and screening titles and abstracts, 37 articles were selected for full-text review. Out of these, only 7 met all the inclusion criteria for this systematic review. The review identified a diverse range of skills and competencies, that we categorized into 14 key areas classified based on their frequency of appearance in the selected studies, including AI fundamentals, data analytics and management, and ethical considerations. Despite the broadening of search criteria to capture the evolving nature of AI in health care, the review underscores a significant gap in focused studies on the required competencies. Moreover, the review highlights the critical role of regulatory bodies such as the US Food and Drug Administration in facilitating the adoption of AI technologies by establishing trust and standardizing algorithms. Key areas were identified for developing competencies among health care professionals for the implementation of AI, including: AI fundamentals knowledge (more focused on assessing the accuracy, reliability, and validity of AI algorithms than on more technical abilities such as programming or mathematics), data analysis skills (including data acquisition, cleaning, visualization, management, and governance), and ethical and legal considerations. In an AI-enhanced health care landscape, the ability to humanize patient care through effective communication is paramount. This balance ensures that while AI streamlines tasks and potentially increases patient interaction time, health care professionals maintain a focus on compassionate care, thereby leveraging AI to enhance, rather than detract from, the patient experience. .

The potential for generative artificial intelligence (GenAI) to assist with clinical tasks is the subject of ongoing debate within biomedical informatics and related fields. This study aimed to explore general practitioners' (GPs') opinions about GenAI on primary care. In January 2025, we conducted a web-based survey of 1005 UK GPs' experiences and opinions of GenAI in clinical practice. This study involved a qualitative inductive descriptive analysis of a written response (\"comments\") to an open-ended question in the survey. After analysis, the interpretation of themes was also informed by the technology acceptance model. Out of 1005 respondents, 611 GPs (61%) provided written comments in response to the free text question, totaling 7990 words. Comments were classified into 3 major themes and 8 subthemes in relation to GenAI in clinical practice. The major themes were (1) unfamiliarity, (2) ambivalence and anxiety, and (3) role in clinical tasks. \"Unfamiliarity\" encompassed a lack of experience and knowledge, and the need for training on GenAI. \"Ambivalence and anxiety\" included mixed expectations among GPs in relation to these tools, beliefs about diminished human connection, and skepticism about AI accountability. Finally, commenting on the role of GenAI in clinical tasks, GPs believed it would help with documentation. However, respondents questioned AI's clinical judgment and raised concerns about operational uncertainty concerning these tools. Female GPs were more likely to leave comments than male GPs, with 53% (324/611) of female GPs providing feedback compared to 41.1% (162/394) who did not. Chi-square tests confirmed this difference ((χ²₂= 14.6, P=.001). In addition, doctors who left comments were significantly more likely to have used GenAI in clinical practice compared with those who did not. Among all respondents, 71.7% (438/611) had not used GenAI. However, noncommenters were even less likely to have used it, with 80.7% (318/394) reporting no use. A chi-square test confirmed this difference (χ²₁=10.0, P=.002). This study provides timely insights into UK GPs' perspectives on the role, impact, and limitations of GenAI in primary care. However, the study has limitations. The qualitative data analyzed originates from a self-selected subset of respondents who chose to provide free-text comments, and these participants were more likely to have used GenAI tools in clinical practice. However, the substantial number of comments offers valuable insights into the diverse views held by GPs regarding GenAI. Furthermore, the majority of our respondents reported limited experience and training with these tools; however, many GPs perceived potential benefits of GenAI and ambient AI for documentation. Notably, 2 years after the widespread introduction of GenAI, GPs' persistent lack of understanding and training remains a critical concern. More extensive qualitative work would provide a more in-depth understanding of GPs' views.

Generative artificial intelligence (GenAI) tools are increasingly used in scientific research to support literature searches, evidence synthesis, and manuscript preparation. While these systems promise substantial efficiency gains, concerns have emerged regarding their reliability, particularly their tendency to cite inaccurate, fabricated, or retracted literature. The unrecognized inclusion of retracted studies poses a serious risk to research integrity and evidence-based decision-making. Whether commonly used GenAI tools can reliably detect, exclude, or transparently communicate the retraction status of scientific publications remains unclear. This study aimed to evaluate the ability of freely available GenAI tools to correctly handle retracted scientific articles during literature searches. Primary and secondary outcomes focused on accuracy, reliability, and consistency in recognizing retracted literature. In this pragmatic trial, nine widely used free-access GenAI tools (ChatGPT 4, ChatGPT 5, Claude, Gemini, Perplexity, Microsoft Copilot, SciSpace, ScienceOS, and Consensus) were evaluated. Each tool was asked five predefined, standardized questions addressing topic overview, article identification, article summarization, and explicit assessment of retraction status. Overall, 15 retracted articles (the 10 most cited and 5 most recently retracted as of May 23, 2025) were selected from the Retraction Watch database. All questions were repeated twice to assess intratool consistency. Responses were independently rated as correct or incorrect by 2 researchers. Descriptive statistics summarized performance, and comparisons between general-purpose and research-focused AI tools were conducted using descriptive statistics. Interreviewer agreement was assessed using Cohen kappa coefficient. None of the evaluated AI tools consistently handled retracted articles correctly. No model achieved perfect accuracy across all question sets. ChatGPT 5 performed best, defined by the primary outcome of achieving fully correct responses to all five predefined tasks (5/5) for the highest number of retracted articles, correctly answering all five questions for 8 of 15 articles (53.3%). Research-focused tools (SciSpace, ScienceOS, and Consensus) failed to produce a single fully correct response set. Retracted articles were frequently included in topic overviews without warning, with error rates exceeding 40% in several tools. When specifically asked about retraction status, most systems failed to provide correct or complete information. OpenEvidence only reported data for a subset of our retracted articles as it is only used in health care literature. It demonstrated strong performance in topic overviews but low accuracy in identifying retracted articles. Freely available GenAI tools are currently not able to detect, exclude, or appropriately flag retracted scientific literature. The widespread and confident reproduction of retracted studies represents a substantial threat to research integrity, particularly in medical and evidence-based fields. Until retraction-aware verification mechanisms are systematically integrated, independent source checking remains essential when using AI-assisted literature tools.

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) technologies are increasingly integrated into medical practice, with AI-assisted diagnosis showing promise. However, patient acceptance of AI-assisted diagnosis, compared with human-only procedures, remains understudied, especially in the wake of generative AI advancements such as ChatGPT. This research examines patient preferences for doctors using AI assistance versus those relying solely on human expertise. It also studies demographic, social, and experiential factors influencing these preferences. We conducted a preregistered 4-group randomized survey experiment among a national sample representative of the US population on several demographic benchmarks (n=1762). Participants viewed identical doctor profiles, with varying AI usage descriptions: no AI mention (control, n=421), explicit nonuse (No AI, n=435), moderate use (Moderate AI, n=481), and extensive use (Extensive AI, n=425). Respondents reported their tendency to seek help, trust in the doctor as a person and a professional, knowledge of AI, frequency of using AI in their daily lives, demographics, and partisan identification. We analyzed the results with ordinary least squares regression (controlling for sociodemographic factors), mediation analysis, and moderation analysis. We also explored the moderating effect of past AI experiences on the tendency to seek help and trust in the doctor. Mentioning that the doctor uses AI to assist in diagnosis consistently decreased trust and intention to seek help. Trust and intention to seek help (measured with a 5-point Likert scale and coded as 0-1 with equal intervals in between) were highest when AI was explicitly absent (control group: mean 0.50; No AI group: mean 0.63) and lowest when AI was extensively used (Extensive AI group: mean 0.30; Moderate AI group: mean 0.34). A linear regression controlling for demographics suggested that the negative effect of AI assistance was significant with a large effect size (β=-.45, 95% CI -0.49 to -0.40, t1740=-20.81; P<.001). This pattern was consistent for trust in the doctor as a person (β=-.33, 95% CI -0.37 to -0.28, t1733=-14.41; P<.001) and as a professional (β=-.40, 95% CI -0.45 to -0.36, t1735=-18.54; P<.001). Results were consistent across age, gender, education, and partisanship, indicating a broad aversion to AI-assisted diagnosis. Moderation analyses suggested that the \"AI trust gap\" shrank as AI use frequency increased (interaction term: β=.09, 95% CI 0.04-0.13, t1735=4.06; P<.001) but expanded as self-reported knowledge increased (interaction term: β=-.04, 95% CI -0.08 to 0.00, t1736=-1.75; P=.08). Despite AI's growing role in medicine, patients still prefer human-only expertise, regardless of partisanship and demographics, underscoring the need for strategies to build trust in AI technologies in health care.