When AI Helps and Harms at the Same Time
Clinical Decision Support Systems (CDSS) are AI-powered tools designed to assist medical professionals in clinical decision-making from analyzing patient data and generating diagnostic recommendations to providing therapy guidance based on the latest clinical evidence. These systems have demonstrably improved workflow efficiency and accelerated decision-making in hospital settings.
Yet behind those benefits, the researchers identified a serious threat known as automation bias a condition in which medical personnel automatically accept system recommendations without performing critical verification. As doctors grow accustomed to depending on AI output, the diagnostic reasoning process that should involve deep analysis, clinical experience, and professional reflection is quietly eroded.
How the Research Was Conducted
The research team conducted a systematic narrative literature review drawing from international scientific databases including Scopus, Web of Science, PubMed, and Google Scholar. Articles were limited to publications from 2020 to 2024, covering empirical studies, systematic reviews, and conceptual articles that examine the interaction between healthcare workers and AI systems in clinical decision-making contexts.
Through that synthesis process, the researchers identified four major themes describing the dynamics of cognitive dependency on CDSS.
Four Key Findings
- A Shift in the Locus of Clinical Evaluation Intensive CDSS use shifts the evaluation process away from a doctor's internal analysis toward external, algorithm-based validation. Rather than independently constructing diagnostic hypotheses, clinicians tend to confirm the system's output. At high levels of dependency, reflective engagement in the differential diagnosis process declines significantly.
- Automation Bias Intensifies Under Work Pressure In clinical environments with high time pressure and heavy workloads, medical staff increasingly lean on system recommendations. More alarmingly, when the system produces inaccurate recommendations, doctors with high levels of dependency tend to overlook clinical indicators that contradict the AI's output.
- The Dual Role of CDSS as a Cognitive Augmentation Tool CDSS holds genuine potential to expand clinical analytical capacity extending working memory, providing rapid access to up-to-date medical literature, and supporting the recognition of complex patterns from large datasets. However, the effectiveness of these augmentative functions depends heavily on system design. Systems that include logical explanations (explainable AI) encourage analytical engagement, while systems that display only a final result without explanation promote passive acceptance.
- A Transformation of Trust in Human–AI Interaction Trust in CDSS is not built solely on technical accuracy it is also shaped by institutional legitimacy. When a system is formally endorsed by a healthcare institution, clinicians tend to internalize the system's authority as part of standard practice, which in turn deepens cognitive dependency if no reflective mechanism is in place.
The Deskilling Risk: Doctors Who Lose Their Edge
One of the most serious implications of these findings is the phenomenon of deskilling a decline in clinical ability caused by reduced independent cognitive training. Intensive CDSS use without a reflection mechanism reduces opportunities for doctors to independently exercise their differential diagnosis skills, fostering a thinking pattern that is more reactive to technology recommendations than proactively clinical.
Zahratul Hayati and her team stress that the impact of CDSS on clinical decision quality is non-linear. At low to moderate levels of use, the system contributes positively to efficiency and accuracy. But once dependency exceeds a certain threshold, the integrity of diagnostic reasoning becomes compromised.
The Solution: A Balanced Human–AI Collaboration Model
The researchers advocate for a human-centered AI approach in CDSS implementation positioning technology as a cognitive partner, not a replacement for professional judgment. In practical terms, this means:
- CDSS should be designed to present alternative diagnoses and levels of recommendation uncertainty, not just a single final answer
- Explainable AI features should be a system standard, not an optional add-on
- Healthcare institutions need to develop usage guidelines that encourage independent verification by clinicians
- Medical education and continuing professional development curricula must integrate AI literacy and automation bias mitigation training
Author Profiles
Zahratul Hayati is a researcher and academic at Akademi Kebidanan Surya Mandiri Bima, focusing on digital health and technology-based clinical decision-making.
Imran Yaman is affiliated with STIKES Marendeng Majene, and is actively engaged in health informatics and medical information systems research.
Yayah Sya'diah of Akademi Perekam Medis dan Infokes Bhumi Husada specializes in medical records and health information management.
Research Source
Article Title: Cognitive Dependency on Clinical Decision Support Systems: Implications for Diagnostic Reasoning Journal: Asian Journal of Healthcare Analytics (AJHA), Vol. 5, No. 1, 2026, pp. 59–70
0 Komentar