Interdisciplinary group proposes guidelines for the use of AI in science

Artificial intelligence (AI) produces text, videos and images that are nearly indistinguishable from human ones, often leaving us confused about what is real. Researchers and scientists are increasingly being supported by AI. That's why an international task force has developed principles for the use of AI in research to ensure trust in science.

Science thrives on reproducibility, transparency, and accountability, and trust in research derives, among other things, from the fact that results are valid regardless of the institution in which they were obtained. Furthermore, the data underlying the research must be made publicly available, and researchers must be held accountable for their publication.

But what happens when AI gets involved in research? Experts have long used AI tools to design new molecules, evaluate complex data, generate research questions, and make mathematical inferences. I even had to prove it. AI is changing the face of research, and experts debate whether its results can still be trusted.

According to an editorial in the latest issue of the journal, an interdisciplinary working group made up of members from politics, business and academia, five principles should continue to be observed to ensure human responsibility in research. Proceedings of the National Academy of SciencesUrs Gasser, professor of public policy, governance and innovative technologies at TUM, is one of the experts.

A quick summary of our recommendations is as follows:

• Researchers must disclose the tools and algorithms they used and clearly identify machine and human contributions.
• Even when using AI analysis tools, researchers remain responsible for the accuracy of their data and the conclusions drawn from it.
• AI-generated data needs to be labeled so that it cannot be confused with real-world data or observations.
• Experts need to ensure that their findings are scientifically sound and do no harm – for example, there should be minimal risk that the AI ​​will be “biased” by the training data used.
• Finally, researchers should work with policymakers, civil society, and businesses to monitor the impact of AI and adapt methods and rules as needed.

“Previous AI principles were primarily concerned with the development of AI. The principles currently being developed are focused on scientific applications and have come at the right time. It has a huge impact on researchers in different fields and sectors,” explains Gasser.

The working group recommends that a new strategic council based at the National Academies of Sciences, Engineering, and Medicine should advise the scientific community.

“We hope that other countries' science academies, particularly here in Europe, will pick up this and further stimulate the debate about the responsible use of AI in research,” Gasser says.