Biomedical Ethical Aspects Towards the Implementation of Artificial Intelligence in Medical Education
Open accessAbstract
The increasing use of artificial intelligence (AI) in medicine is associated with new ethical challenges and responsibilities. However, special considerations and concerns should be addressed when integrating AI applications into medical education, where healthcare, AI, and education ethics collide. This commentary explores the biomedical ethical responsibilities of medical institutions in incorporating AI applications into medical education by identifying potential concerns and limitations, with the goal of implementing applicable recommendations. The recommendations presented are intended to assist in developing institutional guidelines for the ethical use of AI for medical educators and students.
AI snapshot
Written from the paper’s full text.
TL;DR
This commentary argues that medical institutions should establish a biomedical ethical framework—built on the four principles of autonomy, justice, non-maleficence, and beneficence—before integrating AI applications into medical education. It offers concrete recommendations for institutional guidelines aimed at medical educators and students.
Key findings
As a commentary, the paper advances recommendations rather than empirical results:
- On autonomy: institutions should ensure AI use is transparent and comprehensible, obtain users' informed consent, and offer AI as an add-on rather than a replacement for traditional teaching so students can choose whether to use it; LLMs such as ChatGPT "occasionally hallucinate," and dependence on AI may impair development of clinical judgment.
- On justice: the paper cites an estimate that 60% of data used to develop AI applications will be synthetic by 2024 while 40% will still be real data, raising questions of how to compensate and acknowledge data sources; it recommends equity frameworks, financial assistance/scholarships, diverse training data, and regular audits for bias.
- On non-maleficence: AI algorithms must be designed, validated, and evaluated to avoid misleading users; users should be educated about functionalities, known biases, and risks; privacy regulations (GDPR, HIPAA) and approvals (FDA, EMA, institutional review boards) should be observed, and individual performance data handled confidentially.
- On beneficence: appropriate training (tutorials, supervised workshops, online/offline materials) before implementation, collaboration with all stakeholders, and regular assessment/optimization of AI applications maximize benefits.
Methods
Design: a commentary (not an empirical study) that identifies potential concerns and limitations of AI in medical education and proposes recommendations within a biomedical ethics framework. Data/sample: none; the work synthesizes prior literature and cites 51 references. Analysis: conceptual discussion organized around the four principles of biomedical ethics (autonomy, justice, non-maleficence, beneficence), summarized in a figure of key recommendations.
Limitations
The paper does not state explicit limitations. Evident from the text: it is a commentary offering normative recommendations without empirical validation or implementation data; several recommendations (e.g., verifying training data, controlling data-sharing agreements with external vendors) are acknowledged as challenging to apply in practice, particularly for closed-source or externally developed AI products.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 25 cited sources checked · 0 supported · 0 flagged
- Interpretationreasoning
Integrating AI into medical education poses significant challenges to the autonomy of users, including students, educators, and medical professionals.
“However, integrating AI into medical education poses significant challenges to the autonomy of users, including students, educators, and medical professionals.”
Quote in the paperSubstantial claim - Premisecitation
The dependence on AI technology may result in the lack of development of essential decision-making skills and clinical judgment.
“The dependence on AI technology may also result in the lack of development of essential decision-making skills and clinical judgment [29].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [29] Amann J, Blasimme A, Vayena E, et al. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak. 2020;20:310. https://doi.org/10.1186/s12911-020-01332-6. source
Not checked in the free run.
- Premisecitation
There is a risk that AI systems will compromise patient autonomy and dignity by making treatment decisions without appropriate oversight.
“Finally, there is a risk that AI systems will compromise patient autonomy and dignity by making treatment decisions without appropriate oversight [15, 20].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [15] Rigby MJ. Ethical dimensions of using artificial intelligence in health care. AMA J Ethics. 2019;21(2):121–4. https://doi.org/10.1001/amajethics.2019.121. source
Not checked in the free run.
- cites [20] Wilhelm D, Hartwig R, McLennan S, et al. Ethical, legal and social implications in the use of artificial intelligence-based technologies in surgery: principles, implementation and importance for the user. Chirurg. 2022;93(3):223–33. https://doi.org/10.1007/s00104-022-01574-2. source
Not checked in the free run.
- Premisecitation
Medical students who have received training in AI may be better equipped and more comfortable using AI tools and technologies in their later clinical work.
“Moreover, medical students who have received training in AI may be better equipped and more comfortable using AI tools and technologies in their later clinical work as they learn to understand and use AI concepts and their potential applications from the beginning of their careers [12, 13].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [12] Sit C, Srinivasan R, Amlani A, et al. Attitudes and perceptions of UK medical students towards artificial intelligence and radiology: a multicentre survey. Insights Imaging. 2020;11(1):14. https://doi.org/10.1186/s13244-019-0830-7. source
Not checked in the free run.
- cites [13] Masters K. Artificial intelligence in medical education. Med Teach. 2019;41(9):976–80. https://doi.org/10.1080/0142159X.2019.1595557. source
Not checked in the free run.
- Premisecitation
Advanced computational models for natural language processing, such as ChatGPT, occasionally hallucinate, generating information or responses that are not based on factual data.
“For example, advanced computational models for natural language processing, such as ChatGPT, occasionally hallucinate, i.e., generate information or responses that are not based on factual data [28].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [28] Dziri N, Milton S, Yu M, Zaiane O, Reddy S. On the origin of hallucinations in conversational models: Is it the datasets or the models? arXiv. 2022;arXiv:220407931. https://doi.org/10.48550/arXiv.2204.07931. source
Not checked in the free run.
- Premisecitation
Current publications on AI-related medical education emphasize key competencies or explore educational AI programs and concepts, but fail to specifically account for the four main pillars of biomedical ethics: autonomy, justice, non-maleficence, and beneficence.
“While current publications on AI-related medical education emphasize key competencies or explore educational AI programs and concepts, they fail to specifically account for the four main pillars of biomedical ethics: autonomy, justice, non-maleficence, and beneficence [21–24].”
Quote in the paperSubstantial claim0 of 4 sources checked- cites [21] Beauchamp TL, Childress JF. Principles of biomedical ethics. 8th ed. Oxford: Oxford Publishing Press; 2019. p. 1–512.
Not checked in the free run.
- cites [22] Varkey B. Principles of clinical ethics and their application to practice. Med Princ Pract. 2021;30:17–28. https://doi.org/10.1159/000509119. source
Not checked in the free run.
- cites [23] Çalışkan SA, Demir K, Karaca O. Artificial intelligence in medical education curriculum: an e-Delphi study for competencies. PLoS ONE. 2022;17(7):e0271872. https://doi.org/10.1371/journal.pone.0271872. source
Not checked in the free run.
- cites [24] Charow R, Jeyakumar T, Younus S, et al. Artificial intelligence education programs for health care professionals: scoping review. JMIR Med Educ. 2021;7(4):e31043. https://doi.org/10.2196/31043. source
Not checked in the free run.
- Premisecitation
LLMs could be used to cheat on exams, papers, and assignments.
“While LLMs can explain medical terminology to students (and patients) or simulate anamnestic interviews, they could also be used to cheat on exams, papers, and assignments [14].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [14] Kung TH, Cheatham M, Medinilla A, et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. medRxiv. 2022. https://doi.org/10.1101/2022.12.19.22283643. source
Not checked in the free run.
- Premisecitation
Constantly changing applications or expanded use cases may lead to newly identified risks or harms associated with AI, such as privacy or security concerns, discriminatory outcomes, and embedded or inserted biases.
“However, constantly changing applications or expanded use cases may also lead to newly identified risks or harms associated with AI, such as privacy or security concerns, discriminatory outcomes, and embedded or inserted biases [49].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [49] Zhou J, Chen F, Berry A, Reed M, Zhang S, Savage S, editors. A survey on ethical principles of AI and implementations. IEEE SSCI. 2020;1–4. https://doi.org/10.1109/SSCI47803.2020.9308437. source
Not checked in the free run.
- Premisecitation
It is often not possible to obtain complete and detailed information about training data for medical AI applications, for example, when applications are not open source or when knowledge about the technology used is missing.
“However, this is often not applicable, for example, when applications are not open source or with missing knowledge about the technology used [43].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [43] Bellamy RKE, Dey K, Hind M, et al. AI Fairness 360: an extensible toolkit for detecting and mitigating algorithmic bias. IBM J Res Dev. 2019;63(4/5):4:1–4:15. https://doi.org/10.1147/JRD.2019.2942287. source
Not checked in the free run.
- Premisecitation
AI algorithms trained on biased or incomplete data can lead to inaccurate or discriminatory outcomes and potential harm to specific patient groups, such as women.
“This may lead to inaccurate or discriminatory outcomes and potential harm to specific patient groups, such as the recently discovered bias against women in AI tools [42].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [42] Dastin J. Amazon scraps secret AI recruiting tool that showed bias against women. In: Martin K, editor. Ethics of Data and Analytics. 1st ed. Boca Raton, FL: CRC Press; 2018. p. 296–9.
Not checked in the free run.
- Premisecitation
The complexity and opacity of AI algorithms can make it challenging to comprehend how algorithms arrive at specific decisions, potentially reducing the ability to make informed judgments about the appropriateness and correctness of AI-generated recommendations or diagnoses.
“Moreover, the complexity and opacity of AI algorithms can make it challenging to comprehend how algorithms arrive at specific decisions, potentially reducing the ability to make informed judgments about the appropriateness and correctness of AI-generated recommendations or diagnoses [30].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [30] Durán JM, Jongsma KR. Who is afraid of black box algorithms? On the epistemological and ethical basis of trust in medical AI. J Med Ethics. 2021;47:329–35. https://doi.org/10.1136/medethics-2020-106820. source
Not checked in the free run.
- May combine two claims. 8
- Classified as probably not a substantial claim. 4
May combine two claims.
PremisecitationLLMs can explain medical terminology to students (and patients) or simulate anamnestic interviews.
“While LLMs can explain medical terminology to students (and patients) or simulate anamnestic interviews, they could also be used to cheat on exams, papers, and assignments [14].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [14] Kung TH, Cheatham M, Medinilla A, et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. medRxiv. 2022. https://doi.org/10.1101/2022.12.19.22283643. source
Not checked in the free run.
May combine two claims.
PremisecitationFurther concerns include the insufficient protection of data privacy and confidentiality, as well as the lack of informed consent when retrospectively using patient data for training.
“Further concerns include the insufficient protection of data privacy and confidentiality, as well as the lack of informed consent when retrospectively using patient data for training [18, 19].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [18] Luxton DD. Recommendations for the ethical use and design of artificial intelligent care providers. Artif Intell Med. 2014;62(1):1–10. https://doi.org/10.1016/j.artmed.2014.06.004. source
Not checked in the free run.
- cites [19] Kargl M, Plass M, Müller H. A literature review on ethics for ai in biomedical research and biobanking. Yearb Med Inform. 2022;31(1):152–60. https://doi.org/10.1055/s-0042-1742516. source
Not checked in the free run.
May combine two claims.
InterpretationreasoningAI applications may perpetuate existing inequities in the medical field by providing unequal access to care or making biased treatment decisions.
“Hence, AI applications may perpetuate existing inequities in the medical field by providing unequal access to care or making biased treatment decisions.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationCommon ethical concerns about the use of AI in medicine include lack of transparency, insufficient knowledge about the application used, or false/misleading results.
“Common ethical concerns about the use of AI in medicine include lack of transparency, insufficient knowledge about the application used, or false/misleading results [15, 16].”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [15] Rigby MJ. Ethical dimensions of using artificial intelligence in health care. AMA J Ethics. 2019;21(2):121–4. https://doi.org/10.1001/amajethics.2019.121. source
Not checked in the free run.
- cites [16] Sullivan HR, Schweikart SJ. Are current tort liability doctrines adequate for addressing injury caused by AI? AMA J Ethics. 2019;21(2):160–6. source
Not checked in the free run.
May combine two claims.
PremisecitationAI algorithms trained on biased or incomplete data can lead to incorrect diagnoses, unreasonable or unfair decisions, and perpetuate existing inequities in healthcare.
“AI algorithms trained on biased data can lead to incorrect diagnoses and unreasonable or unfair decisions [17].”
- Also stated in: § Non-maleficence · plain
“when AI algorithms are trained based on biased or incomplete data, they can perpetuate or reinforce existing biases and inequities in healthcare.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [17] Chen IY, Szolovits P, Ghassemi M. Can AI help reduce disparities in general medical and mental health care? AMA J Ethics. 2019;21(2):E167-179. https://doi.org/10.1001/amajethics.2019.167. source
Not checked in the free run.
- Also stated in: § Non-maleficence · plain
Probably not a substantial claim: may not pass S2 (checkable), S4 (based).
InterpretationreasoningAppropriate education and training about AI algorithms positively influence beneficence and empower autonomy, justice, and non-maleficence by enabling more informed decision-making, consideration of inequalities and biases, and more effective integration of AI into medical education.
“Moreover, as outlined above, appropriate education and training about AI algorithms not only positively influence beneficence but also empower autonomy, justice, and non-maleficence by enabling more informed decision-making, consideration of inequalities and biases, and more effective integration of AI into medical education.”
Quote in the paperNot a substantial claimMay combine two claims.
PremisecitationWhile retrieving information about drug interactions or dosing regimens using LLMs poses a lower risk, extended use of the models to identify symptoms, diagnosis, and treatment planning without AI-independent verification can be particularly dangerous.
“For example, while simply retrieving information about drug interactions or dosing regimens using LLMs poses a lower risk, extended use of the models to the point of identifying symptoms, diagnosis, and treatment planning without AI-independent verification can be particularly dangerous [45].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [45] WEF. Chatbots RESET. A framework for governing responsible use of conversational AI in healthcare. http://www3.weforum.org/docs/WEF_Governance_of_Chatbots_in_Healthcare_2020.pdfReturn. Accessed 10 Dec 2022.
Not checked in the free run.
Classified as probably not a substantial claim.
Interpretationnone givenAI-generated recommendations may not provide a complete and all-encompassing understanding of a patient's unique medical situation.
“Moreover, students must be aware that AI-generated recommendations may not provide a complete and all-encompassing understanding of a patient’s unique medical situation.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S2 (checkable), S4 (based).
Interpretationnone givenIntegrating AI in medical education has the potential to provide a more immersive, interactive, and personalized learning experience.
“While integrating AI in medical education has the potential to provide a more immersive, interactive, and personalized learning experience, it is essential to establish a biomedical ethical framework beforehand, which should be regularly reevaluated and optimized based on user feedback and the latest developments in the field.”
Quote in the paperNot a substantial claimMay combine two claims.
Premisecitation60% of the data used to develop AI applications is estimated to be synthetic by 2024, and 40% will still be based on real data.
“Although 60% of the data used to develop AI applications is estimated to be synthetic by 2024, 40% will still be based on real data [34].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [34] Castellanos S. Fake it to make it: companies beef up AI models with synthetic data. The Wallstreet Journal. New York, NY: Dow Jones & Company, Inc., 2021. https://www.wsj.com/articles/fake-it-to-make-it-companies-beef-up-ai-models-with-synthetic-data-11627032601. Accessed 15 May 2023.
Not checked in the free run.
May combine two claims.
PremisecitationThe use of AI applications for individual medical education or knowledge acquisition will generate new sensitive data, such as assessment and analysis of individual performance, which may lead to competitive pressures or negative emotions when results are compared with those of fellow students or when faculty members are provided with access.
“The use of AI applications for individual medical education or knowledge acquisition will also generate new sensitive data, for instance, the assessment and analysis of individual performance, which may lead to competitive pressures or negative emotions when results are compared with those of fellow students or when faculty members are provided with access [47, 48].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [47] Graesser AC, D’Mello SK, Strain AC. Emotions in advanced learning technologies. In: Linnenbrink-Garcia L, Pekrun R, editors. International handbook of emotions in education. Abingdon-on-Thames: Routledge; 2014. p. 473–93.
Not checked in the free run.
- cites [48] Anwar M, Greer J. Facilitating trust in privacy-preserving E-learning environments. TLT. 2012;5(1):62–73. https://doi.org/10.1109/TLT.2011.23. source
Not checked in the free run.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
One challenge is that ChatGPT's responses are based on the data on which it was trained, which may result in biased or inaccurate information.
“One challenge is that the responses of ChatGPT are based on the data on which it was trained, which may result in biased or inaccurate information.”
The integration of ChatGPT into medical practices and healthcare settings presents challenges and ethical considerations.
“The integration of ChatGPT into medical practices and healthcare settings presents challenges and ethical considerations.”
Ethical concerns arise when content generated by AI is mistaken for the expertise of a human.
“Ethical concerns arise when content generated by AI is mistaken for the expertise of a human.”
ChatGPT's ability to explain intricate medical terminology in a comprehensible manner aids in knowledge acquisition, fostering continuous learning and improving medical literacy.
“Its ability to explain intricate medical terminology in a comprehensible manner aids in knowledge acquisition, fostering continuous learning and improving medical literacy [6,9,13,14,15,16,17,18,19,20,21,22,23].”
Medical education can benefit significantly from AI-driven tools like ChatGPT.
“Medical education can benefit significantly from AI-driven tools like ChatGPT.”
The citation lists haven’t been fetched yet
References 50 · Cited by 88
Research works printed but not linked: 1
- Aiken R, Epstein R. Ethical guidelines for AI in education: starting a conversation. Int J Artif Intell Educ. 2000;11:163–76.
- Authors
- Busch, F., Adams, L. C., & Bressem, K. K.
- Affiliations
- Humboldt-Universität zu Berlin (DE)
- Stanford Medicine (US)
- Corresponding author
- Busch, Felix
- Venue
- Medical Science Educator
- Published
- 2023-06-07
- Volume
- 33
- Issue
- 4
- Pages
- 1007-1012
- ISSN
- 2156-8650 (ISSN-L)
- PMC
- PMC10403458
- OpenAlex
- W4379793093
- Semantic Scholar
- ebc72fb3653aab0a9b8a1ef5ebfdeec38310dec0
- Cited by
- 88
- References
- 50
- Citations by year
- 2026: 24 · 2025: 19 · 2024: 38 · 2023: 6
- FWCI
- 3.51
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Medical Science Educator (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Refubium (Universitätsbibliothek der Freien Universität Berlin) (submittedVersion, cc-by)
- Universitätsbibliothek der FU Berlin Hochschulschriftenstelle u. Dokumentenserver (cc-by)
- Funders
- Charité - Universitätsmedizin Berlin
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Social Sciences › Safety Research › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
- Keywords
- artificial intelligence in healthcare, medical education, biomedical ethics, AI ethics, institutional guidelines, artificial intelligence in education
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
