Skip to main navigation Skip to main content

CEEM : Clinical and Experimental Emergency Medicine

OPEN ACCESS
ABOUT
BROWSE ARTICLES
FOR CONTRIBUTORS

Articles

Editorial
AI & Digital Health

Artificial intelligence as cognitive support in emergency medicine

Clinical and Experimental Emergency Medicine 2026;13(3):254-257.
Published online: April 3, 2026

Department of Emergency Medicine, Stony Brook University, Stony Brook, NY, USA

Correspondence to: Samita M. Heslin (SamitaHeslin@gmail.com)
• Received: February 27, 2026   • Revised: March 24, 2026   • Accepted: March 24, 2026

Copyright © 2026 The Korean Society of Emergency Medicine

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/).

  • 2,251 Views
  • 28 Download
prev next
Emergency medicine (EM) is a cognitively demanding field in which clinicians perform multiple tasks simultaneously, such as managing a resuscitation, answering a family member's question at the bedside, interpreting a lab or image result, updating a patient's discharge plan, and caring for multiple patients concurrently. The mental work required to accomplish these tasks, such as prioritization, pattern recognition, rapid synthesis of incomplete data, and continuously reassessing, is central to EM [1,2].
Patient complexity, emergency department (ED) crowding, and the demands of the electronic health record (EHR) have added to the workload of emergency clinicians, with growing documentation requirements in the electronic environment associated with decreased professional satisfaction [3–5].
Artificial intelligence (AI) applications are being implemented in multiple areas of EM, including early risk stratification, imaging interpretation, operational forecasting, and clinical decision support [6–9]. Large language models have the potential to support clinical decision-making, streamline workflows, and assist with education and communication in EM [9].
Much of the current discussion is focused on what AI can do, such as the tasks it performs and outcomes it predicts. This editorial asks a different question: what does AI do to cognitive work in EM, and how does it affect the clinician, patient, and trainee? Understanding AI as a form of cognitive support, including where that support fails, is essential to implementing it well. Fig. 1 summarizes the cognitive problems, AI support functions, and key risks across the four dimensions discussed in this editorial.
Human factors engineering focuses on designing systems around how people think and work. This applies directly to the ED, where the effectiveness of an AI system depends not only on how well it works, but also on how it fits into a clinician's workflow [10,11].
EM clinicians work in an environment where time pressure, frequent interruptions, and finite cognitive resources are common. There are multiple design suggestions from human factors engineering that would be beneficial for the integration of AI in EM. Error-tolerant design suggests that if a clinician misreads or misinterprets an AI-generated output, the system should help identify potential errors and facilitate clinician review rather than allow errors to go undetected [10,11]. Designing appropriate feedback mechanisms will help clinicians maintain awareness about what AI systems are and aren't doing. Minimal interaction costs mean AI tools should decrease the number of decision points and steps a clinician has to complete, not increase them [11,12].
In addition to these considerations, human factors engineering and automation literature also describe responsibility drift. When an AI produces a written narrative, risk assessment, or other type of document, there is a lack of clarity about where clinician judgment stops and automated output begins. As a result, clinicians may review documents authored by an AI system with less scrutiny than they would use to review their own rationale [10,12]. Thus, trust calibration, which refers to neither overly relying on nor automatically dismissing AI-generated output, should be developed as a cognitive skill [10,12].
Responsibility drift also raises medicolegal questions. If an AI system generates an inaccurate summary, risk assessment, or discharge instructions, who is responsible? The clinician who reviewed and acted on the AI output, the institution that utilized the AI system, and the developers of the AI system may share some degree of accountability. Furthermore, because of time constraints in the ED, the opportunity to thoroughly review and verify AI-generated content is diminished and therefore increases this concern. While the regulatory and legal frameworks governing AI liability in healthcare continue to evolve, clinicians and institutions utilizing AI systems in clinical environments should understand that AI utilization may not eliminate the clinician's level of responsibility [13,14]. This is an area that calls for further examination as AI adoption in EM grows.
Information in the ED may be located in many areas, such as different EHR tabs, imaging and lab results, prior visit notes, outside records, and consultation notes. EM clinicians often spend meaningful cognitive effort locating and processing information, which has been linked to burnout and dissatisfaction [3–5,15].
AI-driven summarization tools may help with piecing together this fragmented information. By assisting in summarizing diagnoses, hospitalization courses, previous visits, medication changes, and goals of care into concise summaries, these AI tools may reduce search time and return time for direct patient care [16,17]. Large language models using clinical text summarization show early promise in generating coherent and concise summaries, although there is still concern regarding the accuracy and likelihood of hallucinations [17,18].
Documentation support presents a similar opportunity to improve time spent charting in the EHR and lessen cognitive load. AI-assisted and ambient documentation tools have shown early feasibility and potential impact on EHR tasks, although results have been mixed in early studies [8,19]. The ED environment presents some challenges. There may be background noise, other conversations occurring simultaneously, especially during codes when multiple people may be talking at the same time, and care plans changing in real time. Clinical information may be incomplete or develop over time. Therefore, AI-based documentation systems will need to rely on clinician verification to ensure the accuracy and completeness of information. Documentation support should be designed to reflect that diagnoses and clinical impressions often change as new information becomes available [6,8].
When thoughtfully integrated into ED workflows, these AI technologies may help shift cognitive load away from information retrieval and toward clinical reasoning and patient care.
Patient communication is a core component of EM and occurs in every stage of the encounter. For example, explaining differential diagnoses and diagnostic workups, providing reassessments, patient-centered decision-making, discussing disposition decisions, and providing discharge and follow-up instructions. Communicating complex medical information in clear, understandable terms requires cognitive effort, especially in fast-paced encounters where there is little time to repeat or reinforce information. Health literacy may also add to this cognitive burden as medical information should be tailored so it is clear and actionable for each individual patient [20,21].
AI-assisted communication tools may help with part of this cognitive burden. Generative AI has demonstrated potential to create discharge instructions to improve readability and align with health literacy principles [20,21]. When a clinician does not need to construct an explanation from scratch using plain language, that cognitive capacity is free to be used for clinical reasoning. AI-generated materials may be favorably received by patients, particularly in areas such as follow-up guidance and return precautions [20,22].
However, AI-assisted communication tools introduce their own risks. AI tools that produce fluent and confident-sounding language may help reduce the cognitive effort required to explain medical information, but may also introduce other cognitive tasks. In particular, clinicians must assess whether the AI-generated explanation is clinically accurate, does not include hallucinations or fabricated information, and conveys appropriate clinical uncertainty [7,9].
Clinical expertise in EM is often built through cognitive work, including the repeated and often challenging practice of constructing a differential based on early or incomplete clinical information, considering different possible diagnoses, and making decisions before the full clinical picture is clear [23]. AI as cognitive support and its impact on the cognitive development of trainees is an area to examine further.
When AI performs the tasks that would otherwise teach important clinical reasoning skills, such as synthesizing a chart, generating a differential, and making decisions before the diagnosis is clear or test results are back, it may reduce the cognitive tasks that drive learning. Over time, this may create a skills gap and a dependency where clinicians may have difficulty navigating complex situations without AI to help with reasoning [23,24]. Training programs should integrate AI in ways that support cognitive development and independent reasoning, while incorporating AI literacy as an important component of clinical training.
Concretely, training programs could implement structured training exercises in which trainees independently create a differential diagnosis and treatment plan prior to reviewing any AI-generated recommendations. After committing to their own assessment, trainees may compare their work with that of the AI output for review of similarities and differences. Likewise, chart summarizing AI tools may be withheld from the trainee during the initial review of the patient's chart so the trainee can generate their own clinical picture independently, and subsequently use the AI tool to see what information was omitted or misinterpreted. An additional approach is through the use of deliberate AI critique exercises where trainees are presented with an AI-generated output such as a risk assessment, differential diagnosis, or discharge summary. They are then asked to identify errors or instances of overconfidence in the AI output. This educational exercise develops the skills necessary to verify the accuracy of AI-generated outputs as well as to develop the critical thinking skills and calibrated skepticism needed by clinicians when using AI systems. Each of the methods described above preserves the cognitive deliberation that is a component of developing clinical expertise, while also providing trainees with the opportunity to learn about AI systems in an educational setting.
AI implementation in the ED should occur with clinicians, patients, and trainees in mind. While AI may help reduce the cognitive burden on clinicians, it also needs to enable them to easily verify its outputs, support clinician-patient communication without overstating certainty, and provide opportunities for trainee clinical reasoning. The effects of AI on cognitive workload should be assessed through valid workload and attention measures, rather than focusing solely on efficiency and operational impacts [1,10,23,24].
Clinician burnout, documentation burden, and communication gaps are significant challenges. AI has the potential to help address these pressures by reducing the cognitive and administrative demands that compete with it. However, realizing this potential for AI will require consideration not just of how well AI can perform individual tasks, but of how AI redistributes cognitive work across clinicians, patients, and trainees.

Conflicts of interest

Adam J. Singer is the editor-in-chief of this journal, but was not involved in the peer reviewer selection, evaluation, or decision process of this article. The authors have no other conflicts of interest to declare.

Funding

The authors received no financial support for this study.

Data availability

Data sharing is not applicable as no new data were created or analyzed in this study.

Fig. 1.
Artificial intelligence (AI) as cognitive support in emergency medicine: challenges, support functions, and risks. EHR, electronic health record.
ceem-26-104f1.jpg
  • 1. Appelboom Y, Groenen Y, Notten D, et al. Experienced cognitive load in the emergency department. A prospective study. PLoS One 2025;20:e0314052.
  • 2. Weigl M, Müller A, Vincent C, Angerer P, Sevdalis N. The association of workflow interruptions and hospital doctors' workload: a prospective observational study. BMJ Qual Saf 2012;21:399-407.
  • 3. West CP, Dyrbye LN, Shanafelt TD. Physician burnout: contributors, consequences and solutions. J Intern Med 2018;283:516-29.
  • 4. Shanafelt TD, Dyrbye LN, Sinsky C, et al. Relationship between clerical burden and characteristics of the electronic environment with physician burnout and professional satisfaction. Mayo Clin Proc 2016;91:836-48.
  • 5. Asgari E, Kaur J, Nuredini G, et al. Impact of electronic health record use on cognitive load and burnout among clinicians: narrative review. JMIR Med Inform 2024;12:e55499.
  • 6. Kachman MM, Brennan I, Oskvarek JJ, Waseem T, Pines JM. How artificial intelligence could transform emergency care. Am J Emerg Med 2024;81:40-6.
  • 7. Amiot F, Potier B. Artificial Intelligence (AI) and emergency medicine: balancing opportunities and challenges. JMIR Med Inform 2025;13:e70903.
  • 8. Morey J, Jones D, Walker L, et al. Ambient artificial intelligence versus human scribes in the emergency department. Ann Emerg Med 2026;87:561-8.
  • 9. Preiksaitis C, Ashenburg N, Bunney G, et al. The role of large language models in transforming emergency medicine: scoping review. JMIR Med Inform 2024;12:e53787.
  • 10. Carayon P, Hoonakker P, Hundt AS, et al. Application of human factors to improve usability of clinical decision support for diagnostic decision-making: a scenario-based simulation study. BMJ Qual Saf 2020;29:329-40.
  • 11. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Med 2019;17:195.
  • 12. Goddard K, Roudsari A, Wyatt JC. Automation bias: a systematic review of frequency, effect mediators, and mitigators. J Am Med Inform Assoc 2012;19:121-7.
  • 13. Fotheringham K, Smith H. Accidental injustice: healthcare AI legal responsibility must be prospectively planned prior to its adoption. Future Healthc J 2024;11:100181.
  • 14. Mello MM, Guha N. Understanding liability risk from using health care artificial intelligence tools. N Engl J Med 2024;390:271-8.
  • 15. Moy AJ, Schwartz JM, Chen R, et al. Measurement of clinical documentation burden among physicians and nurses using electronic health records: a scoping review. J Am Med Inform Assoc 2021;28:998-1008.
  • 16. Lee C, Vogt KA, Kumar S. Prospects for AI clinical summarization to reduce the burden of patient chart review. Front Digit Health 2024;6:1475092.
  • 17. Van Veen D, Van Uden C, Blankemeier L, et al. Adapted large language models can outperform medical experts in clinical text summarization. Nat Med 2024;30:1134-42.
  • 18. Williams CY, Bains J, Tang T, et al. Evaluating large language models for drafting emergency department encounter summaries. PLOS Digit Health 2025;4:e0000899.
  • 19. Lukac PJ, Turner W, Vangala S, et al. Ambient AI scribes in clinical practice: a randomized trial. NEJM AI 2025;2:10.1056/AIoa2501000
  • 20. Stanceski K, Zhong S, Zhang X, et al. The quality and safety of using generative AI to produce patient-centred discharge instructions. NPJ Digit Med 2024;7:329.
  • 21. Zaretsky J, Kim JM, Baskharoun S, et al. Generative artificial intelligence to transform inpatient discharge summaries to patient-friendly language and format. JAMA Netw Open 2024;7:e240357.
  • 22. Huang T, Safranek C, Socrates V, et al. Patient-representing population's perceptions of GPT-generated versus standard emergency department discharge instructions: randomized blind survey assessment. J Med Internet Res 2024;26:e60336.
  • 23. Natali C, Marconi L, Dias Duran LD, Cabitza F. AI-induced deskilling in medicine: a mixed-method review and research agenda for healthcare and beyond. Artif Intell Rev 2025;58:356.
  • 24. Lyell D, Coiera E. Automation bias and verification complexity: a systematic review. J Am Med Inform Assoc 2017;24:423-31.

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

Include:

Artificial intelligence as cognitive support in emergency medicine
Clin Exp Emerg Med. 2026;13(3):254-257.   Published online April 3, 2026
Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:
Include:
Artificial intelligence as cognitive support in emergency medicine
Clin Exp Emerg Med. 2026;13(3):254-257.   Published online April 3, 2026
Close

Figure

  • 0
Artificial intelligence as cognitive support in emergency medicine
Image
Fig. 1. Artificial intelligence (AI) as cognitive support in emergency medicine: challenges, support functions, and risks. EHR, electronic health record.
Artificial intelligence as cognitive support in emergency medicine