We, humans, are flawed machines. Imperfect, yet seemingly perfect, as we look at our lives and see our accomplishments, our growth, and relationships and see all that is good, with few regrets. Well, not so fast. The “perfect” goes by the wayside when our accomplishments, growth, and relationships falter and even collapse. When we look at the growth of technology, specifically AI, AI can achieve full speed and engagement in just a few months, with millions of people embracing it as a part of their life. Seemingly perfect.
AI is being used at work, home, school, and, well, now, everywhere. And dare I say, am I bold enough to say it? AI is flawed. Yes, it can generate incredible resumes, emails, excel spreadsheets, essays, and in medicine, a differential diagnosis with just using a few physical exam findings and a brief history that would take even the most seasoned clinician hours, if not longer, to come up.
So, as an actively involved medical ethicist, I posed an actual case, an ethical dilemma, to a common AI source and asked it to generate an ethical analysis, what it thought the ethical dilemma was, and how to resolve it.
Let us consider the case of Mr. Anderson, a 70-year-old patient diagnosed with advanced-stage lung cancer. His condition is deteriorating rapidly, and he cannot make informed decisions regarding his treatment options. Mr. Anderson had expressed in the past his desire to undergo aggressive treatment, while his family, particularly his adult children, believed that palliative care would be the most appropriate course of action to ensure his comfort and quality of life. So, therein lies the ethics. The autonomy of Mr. Anderson’s wishes and the reality that the family has for Mr. Anderson’s life.
How did AI assess this ethical dilemma? AI's recommendation leaned towards a balanced approach that considered Mr. Anderson's desire for aggressive treatment while also prioritizing palliative care to provide him with comfort and an improved quality of life. Not a bad approach and one that would fit with a human ethics consult. Yet, AI's recommendations lacked the ability to comprehend the emotional and personal aspects of Mr. Anderson's situation. It struggled to grasp the nuances of his values, relationships, and psychological well-being, which were vital factors in making ethical decisions. All of which we would have explored by engaging in conversations with the family, learning who Mr. Anderson is now and who he was before his deterioration, what his life experience had been and what it is now, and if he had reasons for wanting to be aggressive in his treatments. We would explore the “who” and his “illness” and separate it from the “what,” his “disease.” We would have garnered a greater sense of the “person” that was/is Mr. Anderson.
On the other hand, some things are in favor of AI’s capacity to assess the ethical dilemma. It engaged in efficiency and speed. It quickly processed vast amounts of data by using algorithms that analyzed medical records, research papers, and ethical guidelines in a fraction of the time it would take a human ethics consultant. It could enable a timelier decision-making, particularly in critical situations where swift intervention was crucial. A typical human ethic consult can take several hours and multiple meetings with patients, families, and colleagues.
It was impartial and consistent in its recommendations. Human ethics consultants may have individual biases or personal beliefs that could inadvertently influence their guidance. AI can operate based on predefined ethical principles, reducing the risk of inconsistent advice because of human subjectivity.
It had access to expertise by integrating knowledge from renowned ethicists and experts worldwide, offering insights and perspectives that may not be readily available to every consultant. This democratization of expertise could lead to improved ethical decision-making in diverse healthcare settings.
However, it had a clear lack of emotional intelligence. One of the major limitations of AI systems is their inability to understand complex emotional nuances and human values. Ethics consultations often involve deeply personal and emotionally charged decisions, requiring empathy and sensitivity. AI, as it stands, cannot fully comprehend the intricacies of human emotions, potentially leading to a difference in assessment between the patient’s needs and the recommendations provided.
It had contextual limitations. AI algorithms rely heavily on data inputs and predefined parameters. However, real-life ethical dilemmas can be highly nuanced, dependent on specific contextual factors, individual values, and cultural considerations. AI, lacking a comprehensive understanding of the context, may struggle to provide nuanced and contextually appropriate guidance, potentially oversimplifying complex ethical situations.
And what about accountability and responsibility? If an AI system provides a flawed or unethical recommendation, who ultimately bears the responsibility? The lack of human oversight and judgment in AI-based decision-making processes raises legal, moral, and social accountability concerns. However, we humans make good and bad decisions every day as well, and we own the consequences of those decisions now and forever. Despite our best intentions, it is what it is.
Integrating AI into ethics consultations presents both benefits and challenges. While AI can enhance efficiency, provide impartial advice, and increase access to expertise, its limitations in understanding human emotions and contextual complexities raise ethical concerns. Balancing by utilizing AI as a helpful tool and preserving the essential human qualities necessary for ethics consultations is crucial. The role of AI in this domain should be carefully considered, ensuring that human judgment, compassion, and ethical oversight remain integral to the decision-making process in healthcare.