December 18, 2024

Public Trust in Biomedical Research in the Era of Artificial Intelligence: Opportunities and Challenges

By Susan K. Gregurick, Ph.D., Associate Director for Data Science ar 鶹ý and Director of the Office of Data Science Strategy

Dr. Susan K. Gregurick Dr. Susan K. Gregurick

For many of us, the term “artificial intelligence,” or AI, brings a mix of fascination and fear. AI has immense potential for solving complex biomedical problems. It could reveal harmful mutations in our DNA, find hidden connections in massive data sets, or identify signs of cancers that our human eyes may miss in medical imaging. AI allows for deeper scientific insights and improved health outcomes. However, this new technology needs to be used with care.

Science and health communicators must discuss the opportunities that AI brings for advancing biomedical research and healthcare delivery. But they must also clearly note AI's limitations. AI has made considerable strides in several key areas of biomedical research. Recent breakthroughs include:

  • Protein Structure Prediction and Design: Proteins are involved in nearly every biological process. Many medicines work by interacting with specific proteins in the body. But proteins are long molecules with complex, three-dimensional shapes. Understanding a protein’s precise shape can lead to new and more precise therapies. AI has revolutionized the way we can predict protein structure, and design new ones.
  • Medical Imaging: AI is being used to analyze medical images with exceptional accuracy. This can help radiologists more accurately diagnose conditions such as cancer and cardiovascular diseases.
  • Understanding Complex Conditions: AI can analyze patterns across a broad range of clinical data. This can help researchers identify complex conditions, such as cardiovascular risk, and develop personalized treatments.
  • Analysis of Electronic Health Records (EHRs): Large language models, at type of AI, can sift through vast amounts of medical data at lightning speed to summarize health records, identify trends, and improve patient care.
  • Clinical Decision Support: AI can process vast amounts of patient data and medical knowledge to help healthcare professionals make informed decisions at the bedside with their patients.

AI can be used to quickly process extremely large datasets, identify complex patterns, and generate deeper insights into scientific and medical phenomena. But several critical challenges persist. Chief among them are data quality issues and general AI system reliability—that is, how trustworthy and robust the AI output is. Many datasets are inconsistent or incomplete, and biases in training data can lead to inaccurate or inequitable outcomes. Furthermore, there is a lack of standardization in how biomedical data is collected and processed, which complicates the use of AI systems.

Another challenge that needs to be addressed is the phenomenon of “hallucinations,” in which AI generates misleading or straight-out false information.

The Role of Science Communication

Science communicators must understand both the potential and limitations of AI. Some modern AI systems, including large language models such as ChatGPT and Claude, leverage vast amounts of publicly available information. These principles can help provide a structured approach to communicating about AI in biomedical research:

  • Data validation (the inputs): Science communicators need to ask where the data that primed the AI system came from. Does the data that was fed into the system have inherent biases? For example, was data from only certain types of people?
  • Result verification: Communicators must verify the accuracy and reliability of the AI-generated outputs. That means checking in with both AI and subject matter experts.
  • Ethics: Communicators must remain aware of the ethical concerns around AI. These include data privacy, consent, and true authorship.

Several initiatives have emerged to ensure that AI is used ethically and transparently in biomedical research. In workshops for the biomedical research community, NIH emphasizes the importance of engaging patient communities in AI development and deployment, establishing ethical guidelines and practices, and creating strong partnerships between bioethicists, community participants, and AI developers.

Whether we like it or not, AI is with us now. Science communicators will play an important role in explaining how it works and ensuring its ethical use. The structured approach outlined above to communicating about AI in biomedical research will help to ensure that stakeholders understand the opportunities and challenges in the responsible development and implementation of these powerful technologies.