Global initiative aims to standardize ethical assessments of generative AI in healthcare

In just two months since its launch in 2022, ChatGPT has amassed millions of users worldwide, demonstrating a strong growth in the use of generative artificial intelligence (GenAI) globally. Healthcare users recognize the potential of the rapid growth, evolution, and adoption of these tools, but also face a critical need for a standardized ethical framework.

Led by Duke-NUS Medical School, an international team of researchers examined the existing ethical discourse on GenAI applications in healthcare and developed a pioneering ethics framework check-list to standardize and streamline ethical decision-making.

Ning Yilin, principal investigator and Associate Professor Liu Nan of the Centre for Quantitative Medicine at Duke-NUS Medical School, first and senior authors of the study published in Digital health in The LancetDiscuss your team’s findings and explain why and how the Transparency of Ethics Reporting for Generative AI (TREGAI) Checklist is an essential tool for anyone working in this field.

1. Why was this review of the ethical discourse on the application of GenAI in healthcare necessary?

Although GenAI has interesting potential, ethical concerns have arisen due to its ability to generate realistic textual and visual content, such as medical reports and images. Our motivation was simple: to find out if there were any ethical guidelines or suggestions on the application of GenAI in the healthcare field and to present the available information in a systematic way.

2. How did you conduct the scoping review? What types of documents were you interested in?

We searched databases for English-language articles using search terms related to the concepts “AI ethics,” “generative AI,” and “healthcare.” We were interested in full-length, peer-reviewed research articles that were specific to healthcare and that addressed ethics in the application of GenAI in the health sector.

By analyzing the articles, we examined whether GenAI was the cause of Ethical questionswhether solutions to the problems were proposed and whether they had in-depth discussions about ethics or just brief or general descriptions. We then categorized the ethical issues into nine general ethical principles that were relevant to AI ethical guidelines and the use of AI in healthcare.

3. What are the main conclusions of your study?

We identified four gaps in the current discourse on the ethics and implementation of GenAI in healthcare settings:

• There are no solutions to the ethical issues surrounding the use of GenAI. Regulations and guidelines are insufficient, as interpreting a general ethical principle can be difficult. The complexity of methods and technologies and their rapid evolution can also make it difficult to mitigate ethical issues, even when regulations are in place.
• There is not enough debate about the ethical issues arising from GenAI methods beyond large language models (LLMs) such as ChatGPT. Generative adversarial networks (GANs), used to generate medical research data such as medical images, are one example.
• There is a lack of a common reference for ethical discussions in GenAI research. Most of the papers we reviewed only addressed selected issues, such as privacy. We also found that authors may have different definitions of ethical keywords and may omit or exclude certain keywords without explanation.
• There is not enough discussion about multimodal artificial intelligence. For example, GANs can be used to generate both X-ray images and radiology reports. Due to their complexity and broad capabilities, multimodal artificial intelligence methods can potentially cause more ethical issues if widely implemented.

Having identified these gaps, we felt there was a need for a tool, the TREGAI Checklist, that would enable consistent and effective ethical discussions when implementing GenAI.

4. Tell us a little more about the checklist. How does it fill the four gaps mentioned earlier?

Designed to help users conduct systematic and standardized ethical reviews in research involving GenAI, the checklist is based on nine widely accepted ethical principles – accountability, autonomy, fairness, integrity, confidentiality, security, transparency, and trust – and beneficence, a principle we also consider important.

The checklist encourages transparent documentation of ethical issues and facilitates reviews by ethicists, with whom we encourage users to work for more in-depth ethical assessments.

To our knowledge, our checklist represents the first attempt to create a practical solution to the ethical issues raised in the articles we included in our analysis. While the checklist does not address all of these gaps, it is a tool that can mitigate ethical concerns by guiding users in carrying out comprehensive ethical assessments and evaluations.

5. Who will find the checklist useful?

We developed this checklist primarily for members of the scientific community, such as journal editors, institutional review boards (IRBs), funders, and regulators.

For example, users conducting research using LLM can use it to check whether they have fully considered and addressed the relevant ethical implications of their project. Journal editors can then use it to effectively evaluate submissions for unconsidered ethical issues.

6. Is the checklist only applicable in research-related contexts or can it also be used elsewhere?

The checklist can also be used to assess the benefits, limitations, and risks of other types of GenAI-generated content, including social media posts and educational materials. However, adjustments will need to be made before it can be used in other contexts.

As GenAI adoption becomes more widespread, the checklist may also increase public awareness of the ethical issues involved.

7. Given that GenAI is expected to continue to evolve, how can we be sure that the checklist will remain relevant in the future?

We maintain the TREGAI checklist live online to ensure that we can deploy updates in a timely manner if there are new ethical principles to include changes to recommended action plans and developments in GenAI guidelines and regulations.