Ethics and governance of trustworthy medical artificial intelligence

In the paper

“Data-induced bias refers to the bias when the training data is not representative or insufficient that affects the algorithm model”

What the sources say

• Supported

  The ‘Limitations and open issues’ section states that small or biased training sets can cause overfitting. It also explains that training data unlike the data encountered in practice can lead to poor real-world predictions, supporting the claim’s description of how insufficient or unrepresentative data affect a model. (LLM, conf=0.90)

  cites [27] Mooney SJ, Pejaver V. Big data in public health: terminology, machine learning, and privacy. Annu Rev Public Health. 2018;39(1):95–112. https://doi.org/10.1146/annurev-publhealth-040617-014208. source
• Supported

  The Future challenges section says insufficient data can prevent effective model training. It also says training only on patients with sufficient data introduces selection bias and can make models perform poorly in clinical settings because they do not generalize. (LLM, conf=0.92)

  cites [28] Senders JT, Zaki MM, Karhade AV, et al. An introduction and overview of machine learning in neurosurgical care. Acta Neurochir (Wien). 2018;160(1):29–38. https://doi.org/10.1007/s00701-017-3385-8. source

In the paper

“Healthcare data are sensitive information about a person [23, 83] which is also related to personal privacy.”

What the sources say

• Supported

  The paper’s discussion of medical ethics (page not provided) treats information in patients’ electronic medical records as subject to confidentiality. That supports the broad claim that healthcare data concern personal privacy, even though the paper does not use the word “sensitive.” (LLM, conf=0.85)

  cites [23] Char DS, Shah NH, Magnus D. Implementing machine learning in health care-addressing ethical challenges. N Engl J Med. 2018;378(11):981–3. https://doi.org/10.1056/NEJMp1714229. source
• Supported

  In the 'Data sourcing for MLm must adhere to data protection and privacy requirements' passage, the paper says that data used by medical machine-learning algorithms are subject to privacy protections and discusses the rights of data subjects. This supports the claim that healthcare data concern personal privacy. (LLM, conf=0.91)

  cites [83] Vayena E, Blasimme A, Cohen IG. Machine learning in medicine: addressing ethical challenges. PLoS Med. 2018;15(11):e1002689. https://doi.org/10.1371/journal.pmed.1002689. source

In the paper

“Not only do algorithms inherit human biases, but such biases are also likely to be reinforced and amplified with the accumulation of data and iterations of algorithms”

In the paper

“since most of the texts in TCM are unstructured data and Chinese medicine has more non-standardized expressions, manual annotation is extremely costly and difficult, which greatly affects data quality”

In the paper

“These biases may be continuously replicated and amplified in algorithm and lead to biased prediction results, which may cause discriminatory treatment of certain people in medical care and even lead to medical safety accidents”

In the paper

“The opacity of algorithms affects patients’ and doctors’ trust in medical AI”

In the paper

“In healthcare, biased algorithms may underestimate or overestimate certain patients’ risks, and unrepresentative data may consolidate or exacerbate health disparities”

In the paper

“data quality will directly affect the quality of medical AI algorithm models”

In the paper

“There are three possible reasons for opacity: (1) algorithms are trade secrets that companies intentionally hide; (2) the inability of lay people to understand programming and algorithmic techniques; (3) the complex nature of the algorithms themselves, which are incomprehensible to humans”

What the sources say

• Not checked yetcites [32] Burrell J. How the machine “thinks”: understanding opacity in machine learning algorithms. Big Data Soc. 2016. https://doi.org/10.1177/2053951715622512. source

In the paper

“We adopted a multidisciplinary approach and summarized five subjects that influence the trustworthiness of medical AI: data quality, algorithmic bias, opacity, safety and security, and responsibility attribution”

In the paper

“Algorithmic bias includes both human-induced bias and data-induced bias”

In the paper

“At present, there are some problems in the data of medical AI, such as data errors and omissions in the original data entry process, lack of unified metadata standards, difficulties in data fusion, lack of data management, non-standardized strategy for data cleaning, and much medical data is stored in unstructured forms such as text and images”

In the paper

“Algorithmic bias can affect AI clinical predictions and exacerbate health disparities”

In the paper

“algorithmic errors or security vulnerabilities can pose significant risks and harm to patients”

In the paper

“The involvement of medical AI in clinical practices may threaten doctors ‘and patients’ autonomy and dignity”

In the paper

“deep learning is a typical “black box”, it is opaque and uninterpretable, which makes the biases difficult to be detected”

In the paper

“Medical data are primarily unstructured, lacking uniform and standardized annotation”

In the paper

“Although Watson is good at answering questions, it cannot independently and synthetically evaluate medical problems, nor can it independently conceptualize and search for new relevant information because it lacks the higher level of cognition required for critical thinking and the diversity of solutions ranked by the intrinsic value [67].”

What the sources say

• Not checked yetcites [67] Chung J, Zink A. Hey Watson, can i sue you for malpractice? Examining the liability of artificial intelligence in medicine. Asia Pac J Health Law Eth. 2018;11(2):51–80.

In the paper

“Much medical institution data is stored in the cloud or third-party servers, which is more vulnerable to hacking”

In the paper

“The hybrid approach combines the first two and is the current mainstream of AI algorithm design.”

What the sources say

• Not checked yetcites [77] Wallach W, Allen C, Smit I. Machine morality: bottom-up and top-down approaches for modelling human moral faculties. AI Soc. 2008;22(4):565–82. https://doi.org/10.1007/s00146-007-0099-0. source
• Not checked yetcites [78] Reddy S, Allan S, Coghlan S, et al. A governance model for the application of AI in health care. J Am Med Inform Assoc. 2019;27(3):491–7. https://doi.org/10.1093/jamia/ocz192. source

In the paper

“Medical AI replaces certain tasks previously performed by physicians, which will undoubtedly change the relationship between doctors and patients, and poses a dilemma in the division of ethical responsibilities.”

In the paper

“Beauchamp T.L. and Childress J.F. first proposed four major principles of bioethics in their book Principles of Biomedical Ethics: respect for autonomy, nonmaleficence, beneficence, and justice[76]”

What the sources say

• Not checked yetcites [76] Beauchamp T, Childress J. Principles of biomedical ethics. New York: Oxford University Press; 2013.

In the paper

“In fact, humans have no consensus on ethical dilemmas like the “trolley problem”. It is unlikely to generate an ethical standard that everyone agrees on to deal with the issues, which also poses an ethical challenge for moral algorithms.”

In the paper

“In addition, AI challenges the perceived authority of clinicians and may influence their independent judgment.”

In the paper

“Although people agree with ethical values, there is a huge gap between ethical orientation and application.”

In the paper

“The limited IT power of medical institutions also makes it difficult to ensure data security”

In the paper

“how to find a balance between medical data sharing and patient privacy protection is also an ethical problem faced in medical AI applications”

In the paper

“The value orientation of AI ethics and medical ethics is the same, that is, to promote human health and well-being, and the bottom line is to do no harm to humans.”

In the paper

“The corresponding legal system has not yet been improved, resulting in the lack of effective regulation and restraint in medical data collection, use and privacy protection”

In the paper

“The third type of opacity is inherent to machine learning and is also known as the algorithmic “black box””

In the paper

“It can be seen that, due to the many stakeholders, we cannot simply blame the doctors in the application of medical AI, but need to bring in more responsible stakeholders, such as developers of medical AI, designers of algorithms, providers of training data, regulators of AI, etc.”

In the paper

“Current medical AI is only an auxiliary tool used by doctors to diagnose and treat diseases, not qualified to be a responsible subject.”

In the paper

“Medical AI, as a technology, contains the values of the developer or designer; therefore, those involved have an ethical responsibility for AI products.”

In the paper

“The law is a mandatory norm with certain lagging defects, and the legislative process is characterized by a harshness and long cycle, and more often than not, it can only be “hindsight” and cannot provide timely and effective protection measures.”

In the paper

“medical AI is still just a machine or program at this stage. It cannot adjust its actions to the actual situation and must rely on doctors to manipulate the machinery or make the final decision”

In the paper

“Healthcare data also has high research and business value, but data breach occurs frequently”

In the paper

“It is impossible to develop a code that covers all possibilities. Therefore, safety flaws in AI may endanger more patients than the possibility of a single misdiagnosis by a doctor because automated systems will replicate more errors”

What the sources say

• Not checked yetcites [50] Macnish K. Unblinking eyes: the ethics of automating surveillance. Eth Inf Technol. 2012;14:151–67. https://doi.org/10.1007/s10676-012-9291-0. source

In the paper

“The opacity of the algorithm creates “ignorance” among human agents and will affect the trust of patients and clinicians in AI tools”

In the paper

“This automated setup means that even if the AI gives up control, none of the humans has time to intervene, so who is responsible for the accident?”

In the paper

“From Aristotle's “action must originate from the agent, and a person cannot be unaware of what he or she is doing”, AI neither meets the traditional criteria of free will nor is aware of what it is doing.”

What the sources say

• Not checked yetcites [65] Coeckelbergh M. Artificial intelligence, responsibility attribution, and a relational justification of explainability. Sci Eng Eth. 2020;26(4):2051–68. https://doi.org/10.1007/s11948-019-00146-8. source

In the paper

“In 2022, a report by the National Highway Traffic Safety Administration (NHTSA) showed that in an investigation of 16 Tesla crashes, several showed that Autopilot (Automated Assisted Driving) gave back control of the car to the human driver in “less than a second” on average before the crash and resulted in a total of 15 injuries and one fatality [70].”

What the sources say

• Not checked yetcites [70] Gitlin JM. NHTSA data shows Teslas using Autopilot crashed 273 times in less than a year. 2022. https://arstechnica.com/cars/2022/06/teslas-using-autopilot-crashed-273-times-in-less-than-a-year/. Accessed 28 Jul 2022.

In the paper

“doctors seem to be more repulsed by medical AI than patients. Based on an inherent mindset emphasizing science and logic, doctors are less willing to trust and rely on things they cannot explain”

What the sources say

• Not checked yetcites [40] Feldman R, Aldana E, Stein K. Artificial intelligence in the health care space: how we can trust what we cannot know. Stanf Law Policy Rev. 2019;30:399–419.

In the paper

“Therefore, ethics and morality become an effective complement to the legal system, and AI technology innovation must be carried out in accordance with ethical requirements.”

In the paper

“Some scholars argue that although AI may surpass humans in many aspects, they do not possess free will essentially and does not have moral subjectivity [63, 64].”

What the sources say

• Not checked yetcites [63] Gan SP. How can robots have rights. Eth Res. 2017;3:126–30. https://doi.org/10.3969/j.issn.1671-9115.2017.03.025. source
• Not checked yetcites [64] Sparrow R. Killer robots. J Appl Philos. 2007;24(1):62–77. https://doi.org/10.1111/j.1468-5930.2007.00346.x. source

In the paper

“medical chatbots provide diagnostic and treatment recommendations in order to reduce unnecessary doctor visits. However, these medical chatbots may also harm patients if they are not continuously updated, checked, or regulated”

What the sources say

• Not checked yetcites [7] Katznelson G, Gerke S. The need for health AI ethics in medical school education. Adv Health Sci Educ. 2021;26:1447–58. https://doi.org/10.1007/s10459-021-10040-3. source

In the paper

“The main situations of medical data breaches include cyber attacks such as hacker intrusion and unauthorized access, and even theft or loss of data by insiders in possession of the data”

In the paper

“The data breach is also a security concern”

In the paper

“most current medical AI algorithms are trained using historical data from retrospective studies. When encountering real-world data that differ from that in the training datasets, the performance of AI may be worse, leading to clinical risks”

What the sources say

• Not checked yetcites [52] Kelly CJ, Karthikesalingam A, Suleyman M, et al. Key challenges for delivering clinical impact with artificial intelligence. BMC Med. 2019;17(1):195. https://doi.org/10.1186/s12916-019-1426-2. source

In the paper

“However, it is controversial to grant personality and rights to robots.”

In the paper

“even if a medical AI could perform computer and mathematical-level explanations, it would be difficult to translate them into meaningful medical explanations. It is because AI is always a statistical correlation analysis rather than a medical causality analysis”

What the sources say

• Not checked yetcites [41] Kamensky S. Artificial intelligence and technology in health care: overview and possible legal implications. DePaul J Health Care Law. 2020;21(3):1–13.

In the paper

“The common value goal mentioned in these ethical principles is to put the interests of human beings at the forefront. That is to say, the development and use of AI must promote the good of mankind.”

What the sources say

• Not checked yetcites [71] IEEE. Ethically Aligned Design: A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems, First Edition. 2016. https://ethicsinaction.ieee.org/. Accessed 2 Dec 2022.
• Not checked yetcites [72] Future of life institute. Asilomar AI Principles. 2017. https://futureoflife.org/open-letter/asilomar-ai-principles/. Accessed 2 Dec 2022.
• Not checked yetcites [73] The High-Level Expert Group on AI. Ethics guidelines for trustworthy AI. 2019. https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai. Accessed 2 Dec 2022.

In the paper

“At the beginning of its application, doctors may suffer from machine malfunctions due to inexperience and unskilled operation, some of which (e.g., operating errors in surgical robots) can have serious consequences. Patients would not necessarily have suffered this harm if not for the use of medical AI”

In the paper

“The Turing test suggests that complex AI may have a certain level of consciousness [61].”

What the sources say

• Not checked yetcites [61] Warwick K, Shah H. Passing the turing test does not mean the end of humanity. Cognit Comput. 2016;8(3):409–19. source

In the paper

“The medical AI Watson system passed the U.S. physician licensing exam in 2012 [66].”

What the sources say

• Not checked yetcites [66] Jessica SA. From Jeopardy to Jaundice: the medical liability implications of Dr. Watson and other artificial intelligence systems. La Law Rev. 2013;73:1049.

In the paper

“explaining deep learning-based medical AI is almost impossible, which is related to the connection complexity and mathematical abstraction of deep neural networks. People cannot explain how the machine makes judgments”

What the sources say

• Not checked yetcites [38] Grote T, Berens P. On the ethics of algorithmic decision-making in healthcare. J Med Eth. 2020;46(3):205–11. https://doi.org/10.1136/medethics-2019-105586. source

In the paper

“patients are generally repulsed by unexplained AI interventions in diagnostic and treatment sessions. They usually only accept AI to handle administrative matters such as registration, bill payment, and guidance”

What the sources say

• Not checked yetcites [39] Liu LL, He YM, Liu XD. A survey on patients perceptions and trust in artificial intelligence healthcare. Chin Med Eth. 2019;32(8):986–90. https://doi.org/10.12026/j.issn.1001-8565.2019.08.07. source

In the paper

“Thus, AI involvement may diminish the patient's subject position in clinical practices, undermine shared decision-making between doctors and patients, and threaten patient’s autonomy and dignity.”

In the paper

“The safety issues of medical AI are the risks and harms that occur in its practices, such as program errors, being affected by cybersecurity, the need for adequate testing, difficult software certification, etc., covering various legal and ethical issues”

What the sources say

• Not checked yetcites [47] O’Sullivan S, Nevejans N, Allen C, et al. Legal, regulatory, and ethical frameworks for develop-ment of standards in artificial intelligence(AI) and autonomous robotic surgery. Int J Med Robot. 2019;15(1):e1968. https://doi.org/10.1002/rcs.1968. source

In the paper

“there is a consensus among computer scientists that an inherent tension is between the performance and explainability of machine learning, with the best-performing algorithms being the least transparent and those that provide precise explanatory algorithms often being less accurate”

What the sources say

• Not checked yetcites [46] Holzinger A, Biemann C, Pattichis CS. What Do We Need to Build Explainable AI Systems for the Medical Domain?. arXiv, 2017. https://doi.org/10.48550/arXiv.1712.09923 source

In the paper

“the algorithmic black box makes models lack explainability and are difficult to proofread. If the algorithm is flawed or incorrect, the output will lead to even greater errors, which will most likely cause diagnostic errors, harm human health, and even deprive human lives”

In the paper

“A gap still exists between the machine’s rule-based conditions and probabilistic statistics, and the human judgment mechanism, which is based on emotion and epiphanic meditation.”

In the paper

“The form of the function that relates the input data to the output is too complex for non-specialists to understand, and even the designer may not be able to understand the logic of the calculation”

In the paper

“Only with the participation of all relevant sectors of society and multiple parties can an ethical and publicly acceptable medical AI be developed.”

In the paper

“Trust in AI will improve through a more transparent development process with a better understanding of algorithms and AI functions.”

In the paper

“aspirin was used for about 70 years for its antipyretic, analgesic, and anti-inflammatory clinical effects, but its pharmacological mechanisms were not understood until later [93, 94].”

What the sources say

• Not checked yetcites [93] London AJ. Artificial intelligence and black-box medical decisions: accuracy versus explainability. Hastings Cent Rep. 2019;49(1):15–21. https://doi.org/10.1002/hast.973. source
• Not checked yetcites [94] Vane JR, Botting RM. The mechanism of action of aspirin. Thromb Res. 2003;110(5–6):255–8. https://doi.org/10.1016/S0049-3848(03)00379-7. source

In the paper

“Anonymous data means that it cannot be used to identify a person and is therefore not subject to the GDPR rules, which means that if a company collects anonymous data, it does not need to obtain the consent of the users.”

What the sources say

• Not checked yetcites [86] EU: General Data Protection Regulation. 2018. https://gdpr-info.eu. Accessed 28 Jul 2022.

In the paper

“At the same time, doctors are not exempt from liability because they are the main actors in diagnosing patients.”

In the paper

“Although algorithm transparency is not equal to algorithm explainability, it will form a powerful deterrent and encourage more diverse subjects, such as medical institutions, insurance companies, and social security institutions, to participate in supervision, which will greatly compensate for the lack of supervision of regulatory authorities.”

In the paper

“In practice, however, if software development organizations were to require patient consent for each use of aggregated data, it would inevitably increase the cost of data use.”

In the paper

“Manson and O’Neill [87] argue that more specific consent is not always ethically better and is difficult to achieve in practice. Consent requires unique communicative transactions.”

What the sources say

• Not checked yetcites [87] Manson NC, O’Neill O. Rethinking informed consent in bioethics. Cambridge: Cambridge University Press; 2007. source

In the paper

“From the perspective of human benefits, it is also unethical not to use existing clinical data to develop tools to benefit all humanity [91].”

What the sources say

• Not checked yetcites [91] Larson DB, Magnus DC, Lungren MP, et al. Ethics of using and sharing clinical imaging data for artificial Intelligence: a proposed framework. Radiology. 2020;295(3):675–82. https://doi.org/10.1148/radiol.2020192536. source

In the paper

“In the third scenario, the related people are scrupulous in their duties, but still cannot prevent the medical AI from making an incorrect diagnosis that leads to the patients’ misfortune. There is no clear evidence of who is responsible, or we cannot attribute responsibility to any individual.”

In the paper

“Floridi proposed a principle of moral responsibility of faultless responsibility, which means that no one is at fault, but they are still responsible for it.”

What the sources say

• Not checked yetcites [79] Floridi L, Peng C. Faultless responsibility: on the nature and allocation of moral responsibility for distributed moral actions. Philos Trans R Soc A Math Phys Eng Sci. 2018;374(2083):0112. https://doi.org/10.2139/ssrn.3835211. source

In the paper

“In the previous section, we have elaborated that existing medical AIs are not moral enablers and do not have the ability to think and make decisions independently and cannot be considered as duty bearers.”

In the paper

“However, in either approach, AI’s response to the causality dilemma remains inadequate, and the threat to human moral subjectivity by allowing AI to form universal ethical guidelines spontaneously is even more deadly.”

In the paper

“We believe that it is ethical to dispense with re-informed consent for data use under conditions that ensure data security and do not compromise patient privacy, as long as a sound ethical review system is in place.”

In the paper

“At the legal level, we clarified that medical AI does not have moral status at this stage, and humans remain the responsibility bearers.”

In the paper

“Medicine itself is potentially risky and uncertain.”

In the paper

“At the current level of medical AI development, doctors are still in the position of supervising and they should not let machines make final decisions without their permission.”

In the paper

“For example, operators are required to allow users to express a desire for personal data to be “forgotten”, i.e., “I don’t want you to remember my past data and I want you not to use my data for modeling purposes from now on”. At the same time, the consequences of violating GDPR are severe, and fines can be as high as 4% of the global revenue of the fined organization.”

What the sources say

• Not checked yetcites [86] EU: General Data Protection Regulation. 2018. https://gdpr-info.eu. Accessed 28 Jul 2022.

In the paper

“Patients’ clinical data have potential medical value and should be widely shared to promote the health and well-being of all humans.”

In the paper

“Interpretability is not a necessary or sufficient condition for accountability.”

In the paper

“In 2018, the EU introduced the first bill on personal data privacy protection-General Data Protection Regulation (GDPR) [86]. Unlike previous industry regulations, this is a truly enforceable law with specific and strict requirements.”

What the sources say

• Not checked yetcites [86] EU: General Data Protection Regulation. 2018. https://gdpr-info.eu. Accessed 28 Jul 2022.

In the paper

“For example, China’s Deep Learning Assisted Decision-Making Medical Device Software Approval Points [81] requires quality control of training data, and should ensure diversity of data sources, with data collected from multiple medical institutions at different geographic and hierarchical levels whenever possible. The Approval points further subdivide the data sets into training sets (for algorithm”

What the sources say

• Not checked yetcites [81] Deep Learning Assisted Decision-Making Medical Device Software Approval Points. China National Medical Products Administration [EB/OL] (2019–7–3). https://www.cmde.org.cn/xwdt/shpgzgg/gztg/20190703141714991.html

In the paper

“Macnish and Gauttier [82] argue that it's not appropriate to talk about our relationship with data in terms of ownership. There are only weak philosophical grounds on giving citizens control of ‘their’ data. Control should be based around custody of data and the potential for harm.”

What the sources say

• Not checked yetcites [82] Macnish K, Gauttier S. A pre-occupation with possession: the (non-) ownership of personal data. Big data and democracy. Edinburgh: Edinburgh University Press; 2020. p. 42–56.

In the paper

“Respect for personal privacy is a crucial ethical principle in health care because privacy is linked to personal identity and autonomy [84].”

What the sources say

• Not checked yetcites [84] Cath C. Governing artificial intelligence: ethical, legal and technical opportunities and challenges. Philos Trans A Math Phys Eng Sci. 2018;376(2133):20180080. https://doi.org/10.1098/rsta.2018.0080. source

In the paper

“Current AI technology essentially obtains data by measuring the real world, extracts algorithmic models from the data, and uses the models to make relevant predictions.”

In the paper

“A few companies are keen to propose ethical standards rather than binding rules. The reasons for this are apparent because there is no substantial penalty if they change or disregard ethical standards under this circumstance [80].”

What the sources say

• Not checked yetcites [80] Calo R. Artificial intelligence policy: a primer and roadmap. UC Davis Law Rev. 2017;2:399–436. https://doi.org/10.2139/ssrn.3015350. source

In the paper

“As of 2020, the U.S. Food and Drug Administration (FDA) has approved a total of 222 AI medical device products, and Europe has approved a total of 240 AI medical device products with European conformity certification [97].”

What the sources say

• Not checked yetcites [97] Ujm A, Pd B, Pknv B. Approval of artificial intelligence and machine learning-based medical devices in the USA and Europe (2015–20): a comparative analysis. Lancet Digits Health. 2021;3:195–203. https://doi.org/10.1016/S2589-7500(20)30292-2. source

In the paper

“The main obstacle to data sharing is the ownership of data.”

In the paper

“Some scholars proposed more lenient forms of informed consent, such as broad consent and blanket consent, to facilitate practical implementation [88, 89]. However, the moral rationality for these informed consents remains controversial.”

What the sources say

• Not checked yetcites [88] Virani AH, Longstaff H. Ethical considerations in biobanks: how a public health ethics perspective sheds new light on old controversies. J Genet Couns. 2015;24(3):428–32. https://doi.org/10.1007/s10897-014-9781-9. source
• Not checked yetcites [89] Kaufman D, Bollinger J, Dvoskin R, et al. Preferences for opt-in and opt-out enrollment and consent models in biobank research: a national survey of Veterans Administration patients. Genet Med. 2012;14(9):787–94. https://doi.org/10.1038/gim.2012.45. source

In the paper

“In addition, the algorithms of medical AI are based on the pre-existing human experience.”

In the paper

“However, this distributed responsibility may result in a lack of individual responsibility, leading to a tendency for everyone to be conservative and making the application and innovation of new technologies impossible.”

In the paper

“algorithms may be continually updated beyond their initially approved clinical function, which may require particular policies and supervision.”

In the paper

“Regarding the sharing of health care data, some believe that patients have an obligation to contribute to improving the quality of the health care system [90].”

What the sources say

• Not checked yetcites [90] Faden RR, Kass NE, Goodman SN, et al. An ethics framework for a learning health care system: a departure from traditional research ethics and clinical ethics. Hastings Cent Rep. 2013;43:16–27. https://doi.org/10.1002/hast.134. source

In the paper

“If the doctor has errors in operation, the doctor and the medical institution are responsible.”

In the paper

“Educating users and patients about medical AI is also a way to ensure that they understand the benefits, risks, and limitations of medical AI devices and increase product transparency and user trust [98].”

What the sources say

• Not checked yetcites [98] FDA. Artificial Intelligence and Machine Learning in Software as a Medical Device. 2021. https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device. Accessed 29 Jul 2022.

In the paper

“Besides, current medical AI falls under the category of medical devices”

In the paper

“GDPR requires algorithms to have interpretability, and data subjects can take intervention and require interpretation of the relevant data when they are not satisfied with automated decisions.”

What the sources say

• Not checked yetcites GDPR

In the paper

“If a doctor causes damage, the medical institution can recover compensation from the doctor after taking responsibility.”

In the paper

“algorithmic transparency and traceability do not require algorithms to be explainable, but they provide the possibility of algorithmic explanation and form effective supervision.”

In the paper

“AI robot’s participation in diagnosis and treatment is predicated on the approval of the medical institution where it is located.”

In the paper

“Legal experts concerned with AI governance issues criticize ethical principles as flawed and inadequate in addressing AI’s ethical and social issues.”

What the sources say

• Not checked yetcites [80] Calo R. Artificial intelligence policy: a primer and roadmap. UC Davis Law Rev. 2017;2:399–436. https://doi.org/10.2139/ssrn.3015350. source

In the paper

“Therefore, data and algorithms are the basis of AI computing and decision-making.”