AI Shouldn't Decide Who Dies
The Ethical Concerns of AI in End-of-Life Decisions
Artificial intelligence (AI) is undeniably transforming the field of medicine. From diagnosing diseases to predicting patient outcomes, AI's influence on healthcare is only growing. A quick search on PubMed, a repository of medical research, shows that over 4,000 studies mention ChatGPT, a popular large-language model (LLM). Researchers are exploring the potential of AI for a variety of tasks, such as interpreting pathology slides and responding to patient queries. However, as AI’s presence in healthcare increases, it raises critical ethical questions particularly when it comes to life-or-death situations.
A recent paper published in the Journal of the American Medical Association (JAMA) introduces a controversial idea: using AI as a surrogate in end-of-life conversations. This proposal suggests creating an AI chatbot that would speak on behalf of patients who are incapacitated and unable to communicate their wishes. The AI would gather data from various sources such as social media activity, healthcare decisions, and even religious attendance to predict what the patient might have chosen under specific circumstances. While this concept may sound futuristic, it raises serious ethical concerns. Should AI have the power to make life-and-death decisions for people? As neurosurgeons who frequently engage in end-of-life discussions with patients and their families, we believe this goes too far.
The Emotional Weight of End-of-Life Decisions
End-of-life conversations are some of the most difficult, emotional, and ethically complex discussions in healthcare. Families are often faced with heartbreaking decisions about whether to continue life-sustaining treatment or to allow a loved one to pass peacefully. For physicians, particularly those working in fields like neurosurgery, where traumatic brain injuries, strokes, and tumors are common, these conversations are a regular but challenging aspect of the job.
These moments require more than just medical knowledge they require deep empathy, understanding, and the ability to navigate the raw emotions of the family members involved. Doctors are not just interpreters of clinical data; they are mediators who guide families through some of the darkest moments of their lives. The idea that a chatbot could replace this human connection is not only disheartening but potentially dangerous. AI might be able to process large amounts of data, but it lacks the emotional intelligence and moral judgment that human doctors bring to these critical moments.
Can AI Truly Understand Human Preferences?
The JAMA paper suggests that AI could learn what is important to a patient based on individual-level behavioral data. For example, it might analyze a person’s social media posts, track their travel history, examine donation records, and study past healthcare decisions to predict their preferences. The idea is that this AI model would serve as a proxy for the patient, making decisions based on what it predicts the patient would have wanted.
However, human values, beliefs, and priorities are far more complex than data points. People’s preferences, particularly when it comes to life-and-death decisions, can change based on a variety of factors, including new information, evolving personal circumstances, and emotional states. Moreover, social media posts or past behaviors might not accurately reflect what someone would choose in a specific medical scenario.
What if a patient’s past actions contradict their current preferences?
What if they had a change of heart but never documented it?
A human doctor can take these nuances into account during a conversation; a machine cannot.
The Risks of Delegating Moral Decisions to AI
One of the greatest risks of using AI in end-of-life decisions is that it could inadvertently dehumanize the patient experience. End-of-life care is deeply personal, often shaped by cultural, religious, and family dynamics. The introduction of AI into this space risks reducing the patient’s life to a series of data points, thereby stripping away the emotional and moral complexity of the situation.
Additionally, relying on AI for such decisions could open the door to bias and error. AI systems are only as good as the data they are trained on. If the data is incomplete, biased, or outdated, the AI’s predictions could be flawed. For example, if an AI system is trained primarily on data from a certain demographic group, it may not be able to accurately predict the preferences of patients from different cultural or socioeconomic backgrounds. This could result in decisions that do not align with the patient’s true values or desires.
Another concern is the potential for AI systems to be manipulated or misused. Incompetence or malice could easily infiltrate these systems, leading to decisions that are not in the patient’s best interest. Who would be responsible if the AI makes a wrong call? What if the AI misinterprets the data and recommends a course of action that leads to unnecessary suffering or premature death? The consequences of such mistakes could be devastating.
The Role of the Physician: Empathy and Human Judgment
The heart of the issue lies in the fact that AI, no matter how advanced, lacks the human qualities necessary for making end-of-life decisions. It cannot empathize with a grieving family, understand the weight of cultural or religious values, or offer comfort in a time of despair. End-of-life conversations require more than just technical precision they require compassion, emotional intelligence, and a deep understanding of human nature.
Physicians play a crucial role in these discussions because they bring more than just medical knowledge to the table. They listen to the family’s concerns, answer difficult questions, and help navigate the moral and emotional complexities of the situation. This human connection is irreplaceable, and it’s what makes these conversations so meaningful.
AI can certainly assist physicians in gathering and analyzing data, but it should not replace the human element in end-of-life care. Decisions about life and death require a level of empathy, moral reasoning, and emotional sensitivity that AI simply cannot provide. While AI might be able to predict what a patient "would have wanted," it cannot understand the full scope of what makes us human.
The Future of AI in Medicine: A Tool, Not a Replacement
As AI continues to evolve, it will undoubtedly play an increasingly important role in healthcare. AI has the potential to revolutionize how we diagnose diseases, personalize treatments, and manage healthcare systems. However, it is essential that we remember the limitations of AI, particularly when it comes to ethical and moral decision-making.
AI should be viewed as a tool to support physicians, not as a replacement for human judgment. It can help doctors analyze data more efficiently, identify patterns, and make more informed decisions. But when it comes to deeply personal and ethically charged decisions such as end-of-life care AI should remain in the background, serving only as a supplement to human expertise.
We must also ensure that the use of AI in healthcare is governed by strict ethical guidelines. Physicians and healthcare providers should have the final say in life-and-death decisions, and AI should never be given the authority to make these decisions independently.
Conclusion
Artificial intelligence has the potential to transform the practice of medicine in many positive ways, but there are limits to what it can and should do. AI lacks the emotional intelligence, moral judgment, and empathy necessary to make end-of-life decisions. These deeply personal and complex conversations should remain in the hands of human doctors who can navigate the ethical and emotional dimensions of each unique case.
While AI can assist in gathering and analyzing data, it should never replace the human touch that is so essential in end-of-life care. As we move forward, it is critical that we strike a balance between embracing the benefits of AI and safeguarding the humane aspects of medicine that make us truly human.
