A comprehensive review published in the journal npj Digital Medicine reveals that artificial intelligence (AI) has the potential to enhance the efficiency and accuracy of healthcare professionals. However, these advancements can only be realized when AI systems are intricately designed to align with genuine clinical workflows, foster trust, and establish clear accountability.
The study, titled "Human-AI Collaboration in Healthcare: A Scoping Review," examines recent findings related to the integrations of AI and human professionals within the healthcare sector.
The backdrop of this study underscores the escalating integration of AI into various clinical responsibilities, such as medical record-keeping, task prioritization, triage, image analysis, and overall care coordination.
It's important to note that in high-stakes healthcare environments, the effectiveness of AI cannot simply be gauged by pitting AI systems against the performance of healthcare practitioners. A cooperative approach that ensures significant human oversight is essential in these critical settings, where patient safety, professional accountability, and appropriate decision-making are paramount.
Policy and regulatory bodies such as the World Health Organization (WHO), the European Union’s AI Act, and the U.S. Food and Drug Administration (FDA) stress the need for professional supervision and human oversight when incorporating AI tools in essential healthcare areas to minimize health risks and safeguard individual rights.
In their scoping review, the authors examined recent literature concerning human-AI collaboration. They focused on key aspects such as the effectiveness of AI in different clinical tasks, the technical and human factors contributing to successful collaborations, as well as ethical concerns and governance requirements necessary for accountable AI use.
The review encompassed 140 empirical studies published from January 1, 2015, to October 27, 2025, distilled from an extensive database of 17,463 records. While these studies collectively highlight the advantages of human-AI teamwork in healthcare, the authors caution that comparisons across diverse settings are challenging.
Analysis within three major areas demonstrated that the success of human-AI partnerships is highly dependent on the specific tasks involved. Factors like trust, integration with existing workflows, and adequate training emerged as critical elements influencing successful collaboration. Importantly, a notable discrepancy persists between governance expectations regarding human oversight and the assessments utilized in the reviewed studies.
Specifically, the evaluation of AI's effectiveness varies considerably by task context. For instance, while the effectiveness of collaboration was often measured through immediate task-oriented metrics, broader patient or systemic outcomes were generally overlooked.
In the realm of technical, human, and organizational determinants, the review reveals that effective collaboration is associated with several positive outcomes, such as enhanced performance, increased speed, and improved acceptance, contingent upon the alignment of AI technologies with the workflows and responsibilities of clinical tasks. The most pronounced benefits were found when AI tackled specific, well-defined duties—like case prioritization or content generation—while ensuring that clinicians retained ultimate decision-making authority.
The most robust evidence on human-AI collaboration emerged from studies focused on diagnostic interpretation, whereas evidence for functions like screening and administrative documentation remained less consistent and varied.
Key ethical considerations identified included accountability and patient safety. However, these concerns were often inadequately addressed in the studies’ main evaluations, revealing a gap between policy expectations for oversight and the parameters used in research.
This review underscores the rising significance of human-AI collaboration as a vital approach for the safe and effective adoption of AI systems in healthcare. However, inconsistencies across different tasks, study methodologies, and interpretations of collaboration reflect the need for a more nuanced understanding of these interactions.
The authors advocate for tailored evaluations of collaborative effectiveness that consider the specific contexts and tasks involved. It's crucial to expand metrics beyond mere accuracy and efficiency to include the effects on workflows, cognitive loads, and patient and system outcomes.
Future investigations should also take into account the crucial human and organizational factors—such as trust calibration and interface design—that facilitate successful human-AI cooperation. Systems that empower clinicians to maintain final decision-making roles while applying AI to clearly defined tasks are more likely to achieve beneficial outcomes.
Additionally, the authors emphasize that core ethical principles, particularly accountability and patient safety, should be central to future studies. Merely having human oversight is inadequate without accompanying measures for transparency, accountability, traceability, and defined organizational governance governing AI's role in clinical decisions.
While this scoping review provides valuable insights, the authors acknowledged limitations, including the absence of a formal bias assessment and the exclusive focus on English-language studies and controlled diagnostic interpretation research, which may skew positive results.
Ultimately, these findings lay a crucial foundation for a more task-oriented, longitudinal, and governance-sensitive approach to evaluating human-AI collaboration in healthcare.
