Australian AI ethicist Stefan Harrer, Ph.D., has proposed an ethical framework for the use of Generative AI in healthcare, aiming to address the potential risks and ensure responsible application in the field.
Large Language Models (LLMs), a type of Generative AI, have the power to revolutionize healthcare information management, education, and communication, but they also pose significant dangers if not properly regulated.
Ethical Regulations for AI in Healthcare
Dr. Harrer, who serves as the Chief Innovation Officer of the Digital Health Cooperative Research Centre (DHCRC) and is a member of the Coalition for Health AI (CHAI), emphasizes the need for ethical regulations to govern the development and deployment of generative AI technology.
His study highlights the importance of technical and governance guidance for developers, users, and regulators within the digital health ecosystem to ensure that the potential of generative AI is harnessed safely.
The paper draws attention to various generative AI applications in healthcare, such as assisting clinicians in generating medical reports, simplifying medical jargon for effective clinician-patient communication, and enhancing the efficiency of clinical trial design and drug discovery processes.
However, the study also highlights the dangers associated with LLM-driven generative AI, particularly their ability to produce and disseminate false, inappropriate, and potentially harmful content on a large scale.
Dr. Harrer warns against the hasty release of LLM-powered tools by some entities, which could compromise user well-being and the integrity of AI and knowledge databases.
To address these risks, Dr. Harrer proposes a comprehensive set of risk mitigation pathways tailored to the use of LLM technology in health and medicine.
He calls for an ethical approach to the design and use of generative AI applications, emphasizing the importance of building AI as an assistive tool rather than a replacement for human decision-makers.
The framework also emphasizes transparency, privacy, safety, and performance standards, and advocates for fair work and safe work standards for human developers.
The study suggests alternative approaches to re-designing generative AI applications, shaping regulatory frameworks, and directing research efforts toward implementing and enforcing ethical design and usage principles.
Mitigating AI Risks
Dr. Harrer's proposed regulatory framework comprises ten principles aimed at mitigating the risks associated with generative AI in healthcare.
These principles encompass areas such as human oversight, data transparency, privacy protection, and accountability frameworks for training data and AI-generated content.
The study concludes by highlighting the significance of developing and operating LLM-powered generative AI applications responsibly.
Dr. Harrer foresees a transition from the current competitive environment to a phase of careful experimentation that considers risks, leading to the introduction of specialized applications in digital health data management within the next two years.
Recognizing the crucial role of ethics, stakeholders such as the DHCRC and CHAI acknowledge the need for guidelines and safeguards to ensure the safe and ethical utilization of generative AI.
They stress the importance of protecting patients, reducing bias, and preventing unintended consequences in the field of healthcare.
The full proposed framework was published in EBioMedicine.
Related Article : AI Dives Into the Ocean to Monitor Global Reef Health
