Volume 11, Issue 4 (12-2025)
JMIS 2025, 11(4): 0-0 |
Back to browse issues page
Ethics code: hums
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Mastaneh Z, Mouseli A. From Tool to Clinical Teammate: Redefining AI's Role in Healthcare Teams. , JMIS. 2025; 11 (4)
URL: http://jmis.hums.ac.ir/article-1-618-en.html
URL: http://jmis.hums.ac.ir/article-1-618-en.html
Department of Health Services Management, School of Public Health, Social Determinants in Health Promotion Research Center, Hormozgan Health Research Institute, Hormozgan University of Medical Sciences, Bandar abbas, Iran.
Abstract: (1320 Views)
The integration of artificial intelligence (AI)-based systems into clinical settings, along with their demonstrated performance, has elevated AI from a mere tool to a clinical teammate—an entity that provides information, participates in decision-making, and influences final clinical decisions alongside human professionals [1].
Within this transformative context, physicians and AI systems possess distinct capabilities. Physicians have advantages in understanding the clinical context of patients, their values, clinical judgment, accountability for clinical decisions, and human communication. In contrast, AI excels at processing large volumes of data, identifying complex patterns, and delivering consistent performance. Consequently, their collaborative performance may surpass that of either working independently, generating synergistic benefits for healthcare teams, patients, and healthcare organizations [2].
In a study conducted by Gallo et al. at Stanford Hospital, an AI system based on the Epic Deterioration Index processed vital signs, laboratory results, and electronic health records simultaneously to predict patient deterioration several hours in advance and automatically alert physicians and nurses. This human–AI teaming was associated with a 10.4% reduction in deterioration events, including ICU transfers and rapid response team activations, among high-risk patients. The findings demonstrated how AI can compensate for human limitations in continuous patient monitoring, thereby enhancing the efficiency and safety of clinical teams [3].
Mutual complementarity of human and intelligent elements in clinician–AI teaming
Designing an effective physician–AI team requires an understanding of the strengths and limitations of both actors and the establishment of mutual complementarity, whereby the weaknesses of each can be compensated for by the strengths of the other. Furthermore, their respective blind spots should not substantially overlap; rather, AI algorithms should compensate for areas in which healthcare professionals are prone to error, and vice versa [1, 4]. The relative contribution of AI to teamwork should also vary according to the task, requiring an appropriate balance between human augmentation and automation. AI may assume a more prominent role in routine, high-volume, and standardized activities, whereas physicians should retain the primary role in complex, sensitive, and ethically challenging decisions [4].
The functional distinction between physicians and AI should be designed to avoid legal and ethical ambiguity regarding accountability for clinical decisions. Describing AI as a teammate refers to its role in facilitating decision-making, rather than transferring responsibility. As a technical system, AI lacks legal personhood, moral awareness, and the capacity for independent accountability. Therefore, ultimate responsibility for clinical decisions, even when based on algorithmic outputs, remains with the physician as the legally responsible human decision-maker. This principle is emphasized in contemporary medical ethics literature, international regulatory frameworks, including European Union guidelines on trustworthy AI, and FDA guidance on machine learning-based medical software [5].
Another important aspect of physician–AI team design is the mode of interaction, which is generally classified as sequential or simultaneous. In sequential interaction, physicians first formulate their initial interpretation or decision and subsequently review the AI output to confirm or revise their judgment. In simultaneous interaction, physicians have access to clinical data and AI recommendations from the outset, integrating both information streams into their clinical judgment [6]. This distinction has been illustrated in the context of electrocardiogram (ECG) interpretation in the study by Attia et al. In the sequential approach, physicians independently interpret the ECG and record their diagnosis before reviewing the AI output. In cases of disagreement, they may reconsider relevant electrocardiographic findings. Conversely, in the simultaneous approach, the system displays an AI-generated diagnosis of atrial fibrillation in real time during ECG acquisition, allowing physicians to consider the algorithm's recommendation while examining the waveforms [7].
Co-design and practical explainability in clinician–AI teaming
The interaction between physicians and AI should not be regarded as a purely technical phenomenon focused exclusively on algorithmic accuracy. Rather, it should be examined within a sociotechnical systems framework, which simultaneously considers human, organizational, workflow, and technological factors [8]. Another critical aspect is the involvement of physicians as co-designers in the development of AI systems, which is a prerequisite for successful physician–AI teaming. Active physician participation in the early stages of problem definition, selection of performance metrics, user interface design, and integration of the system into routine workflows enables the consideration of actual clinical needs, workplace constraints, and ethical concerns. Ultimately, this involvement enhances the clinical contextual fit of the system and fosters trust in its use [9].
Another important consideration is the distinction between complete technical transparency and practical explainability in AI-based systems. Rather than focusing solely on revealing the technical details of a model, these systems should provide physicians with highly interpretable, understandable, and predictable outputs. By communicating the degree of uncertainty, AI systems should enable physicians to anticipate system behavior across different clinical situations, monitor its performance, and consciously distinguish their final decisions from algorithmic recommendations when these conflict with clinical judgment. Accordingly, explainability serves as a means of preserving human control, strengthening critical decision-making, and preventing excessive reliance on AI [10].
In summary, a unidimensional focus on the capabilities of AI is insufficient; rather, the respective strengths and limitations of both human and machine must be jointly considered to establish a genuinely complementary system. The successful future of AI in clinical practice lies not in the replacement of physicians, but in the intelligent design of human–AI teams within a socio-technical framework. This approach requires active participation of clinical teams in system design, thoughtful selection of interaction modalities, adherence to principles of practical explainability, and preservation of ultimate human control and accountability. Through such an integrated strategy, the transition from simplistic automation to genuine synergy becomes achievable, enabling the complementary capabilities of physician and machine to meaningfully enhance clinical accuracy, safety, and overall healthcare efficiency.
Within this transformative context, physicians and AI systems possess distinct capabilities. Physicians have advantages in understanding the clinical context of patients, their values, clinical judgment, accountability for clinical decisions, and human communication. In contrast, AI excels at processing large volumes of data, identifying complex patterns, and delivering consistent performance. Consequently, their collaborative performance may surpass that of either working independently, generating synergistic benefits for healthcare teams, patients, and healthcare organizations [2].
In a study conducted by Gallo et al. at Stanford Hospital, an AI system based on the Epic Deterioration Index processed vital signs, laboratory results, and electronic health records simultaneously to predict patient deterioration several hours in advance and automatically alert physicians and nurses. This human–AI teaming was associated with a 10.4% reduction in deterioration events, including ICU transfers and rapid response team activations, among high-risk patients. The findings demonstrated how AI can compensate for human limitations in continuous patient monitoring, thereby enhancing the efficiency and safety of clinical teams [3].
Mutual complementarity of human and intelligent elements in clinician–AI teaming
Designing an effective physician–AI team requires an understanding of the strengths and limitations of both actors and the establishment of mutual complementarity, whereby the weaknesses of each can be compensated for by the strengths of the other. Furthermore, their respective blind spots should not substantially overlap; rather, AI algorithms should compensate for areas in which healthcare professionals are prone to error, and vice versa [1, 4]. The relative contribution of AI to teamwork should also vary according to the task, requiring an appropriate balance between human augmentation and automation. AI may assume a more prominent role in routine, high-volume, and standardized activities, whereas physicians should retain the primary role in complex, sensitive, and ethically challenging decisions [4].
The functional distinction between physicians and AI should be designed to avoid legal and ethical ambiguity regarding accountability for clinical decisions. Describing AI as a teammate refers to its role in facilitating decision-making, rather than transferring responsibility. As a technical system, AI lacks legal personhood, moral awareness, and the capacity for independent accountability. Therefore, ultimate responsibility for clinical decisions, even when based on algorithmic outputs, remains with the physician as the legally responsible human decision-maker. This principle is emphasized in contemporary medical ethics literature, international regulatory frameworks, including European Union guidelines on trustworthy AI, and FDA guidance on machine learning-based medical software [5].
Another important aspect of physician–AI team design is the mode of interaction, which is generally classified as sequential or simultaneous. In sequential interaction, physicians first formulate their initial interpretation or decision and subsequently review the AI output to confirm or revise their judgment. In simultaneous interaction, physicians have access to clinical data and AI recommendations from the outset, integrating both information streams into their clinical judgment [6]. This distinction has been illustrated in the context of electrocardiogram (ECG) interpretation in the study by Attia et al. In the sequential approach, physicians independently interpret the ECG and record their diagnosis before reviewing the AI output. In cases of disagreement, they may reconsider relevant electrocardiographic findings. Conversely, in the simultaneous approach, the system displays an AI-generated diagnosis of atrial fibrillation in real time during ECG acquisition, allowing physicians to consider the algorithm's recommendation while examining the waveforms [7].
Co-design and practical explainability in clinician–AI teaming
The interaction between physicians and AI should not be regarded as a purely technical phenomenon focused exclusively on algorithmic accuracy. Rather, it should be examined within a sociotechnical systems framework, which simultaneously considers human, organizational, workflow, and technological factors [8]. Another critical aspect is the involvement of physicians as co-designers in the development of AI systems, which is a prerequisite for successful physician–AI teaming. Active physician participation in the early stages of problem definition, selection of performance metrics, user interface design, and integration of the system into routine workflows enables the consideration of actual clinical needs, workplace constraints, and ethical concerns. Ultimately, this involvement enhances the clinical contextual fit of the system and fosters trust in its use [9].
Another important consideration is the distinction between complete technical transparency and practical explainability in AI-based systems. Rather than focusing solely on revealing the technical details of a model, these systems should provide physicians with highly interpretable, understandable, and predictable outputs. By communicating the degree of uncertainty, AI systems should enable physicians to anticipate system behavior across different clinical situations, monitor its performance, and consciously distinguish their final decisions from algorithmic recommendations when these conflict with clinical judgment. Accordingly, explainability serves as a means of preserving human control, strengthening critical decision-making, and preventing excessive reliance on AI [10].
In summary, a unidimensional focus on the capabilities of AI is insufficient; rather, the respective strengths and limitations of both human and machine must be jointly considered to establish a genuinely complementary system. The successful future of AI in clinical practice lies not in the replacement of physicians, but in the intelligent design of human–AI teams within a socio-technical framework. This approach requires active participation of clinical teams in system design, thoughtful selection of interaction modalities, adherence to principles of practical explainability, and preservation of ultimate human control and accountability. Through such an integrated strategy, the transition from simplistic automation to genuine synergy becomes achievable, enabling the complementary capabilities of physician and machine to meaningfully enhance clinical accuracy, safety, and overall healthcare efficiency.
Keywords: Clinician-AI Teammate, Complementary, Co-design, Practical explainability, Socio-technical system
Type of Study: Letter to editor |
Subject:
Special
Received: 2026/01/29 | Accepted: 2026/02/22 | Published: 2025/11/1
Received: 2026/01/29 | Accepted: 2026/02/22 | Published: 2025/11/1
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
Contact Us
Post Address: Bandar Abbas, Imam Hossein Boulevard, campus of University of Medical Sciences, School of Nursing and Midwifery.
Tel: +(98) 763 331 7723
Cell Phone: +(98) 939 654 9816
jmis
hums.ac.ir
