LLMs and Generative AI for Healthcare

Chapter 6. Steering the Helm for Ethical Use of LLMs

Ultimately, the goal of using technology in healthcare is to evolve its underlying function to improve outcomes and experiences for patients, clinicians, and all medical practitioners. What goes into designing and delivering healthcare solutions can be radically changed by large language models (LLMs). Their effects could become significant by improving access to medical knowledge and case data, and by mediating between patients and providers.

This starts with a deliberate focus of AI model development in the healthcare context. AI can deliberately curate diverse training datasets relevant to medicine, introduce interpretability features that allow the AI to “open the hood” and explain how it arrived at a decision, and propose algorithmic audits and oversight protocols to ensure the fair treatment of all patient populations. Health value and quality and equitable access to care—not maximal profit or maximal return on investment—should be the goal. There are many articles, papers, and case studies showing the positive effects of AI in healthcare.¹

Ultimately, oversight of trained models in healthcare also needs to adhere to these tenets. This might involve modeling ethical disclosure of what the model can do and cannot do, legislating strict adherence to patient confidentiality and medical ethics, maintaining effective pathways of redress for algorithmic harms, and championing whistleblowing to promote organizational accountability ...