Artificial Intelligence in Biomedical Research and Publications: It is not about Good or Evil but about its Ethical Use

摘要

“From inability to let well alone, from too much zeal for the new and contempt for what is old, from putting knowledge before wisdom, science before art and cleverness before common sense, from treating patients as cases and from making the cure of the disease more grievous than the endurance of the same, good Lord deliver us.” Sir Robert Hutchison (1871 – 1960) The physician’s prayer We are all at different stages of learning, excitement, and usage of the rapidly developing field of artificial intelligence (AI). With increasingly refined algorithms and related applications, AI is progressively getting integrated into contemporary clinical settings as well as research. The inspiration for this editorial was a flyer for an AI event which announced a session: “generate review of literature for thesis using latest AI technologies.” Should that sound as a red flag, or is it one more skilling imperative? We place some of our thoughts here for public health students, researchers, practitioners, and policy makers to engage on this subject and debate the emerging dilemmas. We also need to be aware of the guidelines, regulations, and ethics dimensions of the use of AI in research and publication. We discuss the available national and journal guidelines and point out some gray areas as food for thought for the community medicine and public health fraternity in India. Autonomous AI is a set of advanced system and tools that can perform tasks ranging from basic repetitive activities to analyzing data with limited human oversight and involvement.[1] Debates on addressing human goals for making appropriate autonomy and agency attributions in relation to AI systems revolve around multiple axes including dimensions of ethics and philosophy of action.[2] The Indian Council of Medical Research (ICMR) published the ethical guidelines for application of AI in biomedical research and healthcare in 2023 to provide clarity on liability, transparency, accountability, and oversight.[3] While discussing its potential in improving individual and population health, it cautions us about the ethical, legal, and social concerns in the light of the complexities of algorithmic learning. The ten overarching principles are descriptive and not prescriptive. These are patient or research participant centric, and if followed, can guide stakeholders in using it responsibly. Robust mechanisms that ensure adequate safeguards against stigmatization, discrimination and worsening of vulnerabilities along with secure software designs that minimize risk, maintain privacy and confidentiality of sensitive data are the basis for ethical use of AI in biomedical research. A key concern is trustworthiness, that can only be guaranteed by valid, reliable, and lawful technology which can be reasonably explained by scientific plausibility and transparency. The principle of data privacy includes appropriate security measures, anonymization, and necessary ethical clearance and consent process when using individual data for AI-based algorithms. Other principles include accountability and liability, data quality (free from bias and representative of the target population), accessibility, equity, validity, and inclusiveness in the spirit of collaboration, fairness, and nondiscrimination. These ten principles provide a framework for the entire spectrum of stakeholders—researchers, innovators, product developers, patients, technologists, healthcare professionals, ethicists, and funding agencies. Although these principles are useful in understanding the ethical use of AI in the domain of biomedical science, an emergent aspect is the use of AI in publication practices that needs to be urgently recognized by the scientific community. These issues found a center stage when some scientific publications listed Chat GPT as co-authors.[4] The usage of AI is limited not only to the authors but also extends to its potential use by reviewers and editors. The World Association of Medical Editors (WAM