Pharmacist Brian Spoelhof was helping his institution manage drug shortages during COVID-19 when leaders tapped him for a committee planning how to allocate scarce resources like ventilators.
The experience inspired Spoelhof, director of pharmacy, medication management and systems at UVA Health, to pursue a master’s degree in bioethics at Harvard Medical School. Today, he applies the high-level ethical concepts he learned to one of healthcare’s most significant developments: the use of AI in patient care.
In Ethical Use of AI in Clinical Care and Operational Decision-Making, part of the Joseph A. Oddis Colloquium at the upcoming Pharmacy Futures 2026, Spoelhof joins clinical informatics pharmacist Steven Smoke to discuss the profession’s responsibilities to help ensure ethical, transparent, and patient-centered implementation of AI tools.
The ASHP News Center spoke with Spoelhof to learn more. The following is an edited transcript.
You’ll be giving an overview of different ethical approaches, including utilitarianism and duty-based ethics, to healthcare. Can you tell us a little more about that?
The concept of utilitarianism is very intuitive to healthcare professionals because it is about maximizing benefits. You’re going to act in a way that produces the most benefit for the most amount of people.
For me it was something so intuitive I had never fully considered the drawbacks of that approach. For example, it's a maximized benefit for the most people, not all people. But what are our obligations for that small group of people that we're missing?
This is where duty-based ethics offers an opposing but equally valid theory of what is right and wrong. Rather than assessing an action based on the consequences, we would assess the action and the underlying principles first. In other words, an action can be right even if it doesn’t produce the maximum benefit, or really any benefit for that matter. Though this topic is very complex and nuanced, one way to think about it is to ask, “Do the ends always justify the means?”
What’s an example of that potential conflict?
Let’s take a drug shortage. The utilitarian tries to find the people it's most indicated in and maximize the benefit. But under duty ethics, a person would say, the right thing to do is to give the drug. It doesn't matter if you run out or not. When the person's in front of you, you have to treat them.
Connecting this to AI, you’ll talk about morally disruptive technologies. Can you discuss what that means?
The concept of morally disruptive technology is that some advancements may cause us to reassess how we approach ethical decision making, with or without fundamentally reshaping our core morality. The moral questions we will face with adopting AI aren't necessarily going to be fundamentally new or different, but it will start changing what it means to practice pharmacy. As practice evolves, we may have to start updating our moral norms as we come across new challenges.
What might that look like?
An example in history is full artificial life support. Before we had that, our value would have said death is when your heart stops beating. And with that technology, we had to update what we see as death. Or we may look at clinical decision support as something that is for a clinician to interpret. But with AI, some of that interpretation is being done for you. So what does it mean to come up with clinical decisions?
What's an example of that?
One of the cases that we’ll talk about is vancomycin dosing, something that has long required a pharmacist's expertise. We may even argue that we are morally obligated as pharmacists to do it to ensure safe and effective care. This, however, is a perfect case for AI in patient care. In fact, it's already doing it — possibly better. But if the moral action is the one that promotes the safest and most effective vancomycin treatment, what is the role of pharmacist? We may reconsider our moral norms that a pharmacist be an integral part of vancomycin dosing. We may also have to develop new approaches to what happens if we disagree with the technology.
You also talk about how humans are motivated by moral codes, but AI is fundamentally amoral. Why does that matter in patient care?
At the heart of this is a fundamental question: why do we do what we do? It begins with the recognition that our decisions affect real people. Because we freely choose our actions, we also bear moral responsibility for their consequences, which is why harm can produce genuine regret and moral reflection. This is where AI differs fundamentally from human moral agency. AI does not possess sentience, self-awareness, or an authentic recognition of another human being. Its outputs are inherently computational and probabilistic and therefore it is fundamentally amoral.
This distinction matters because patient care is more than producing answers or completing transactions; it is a human act grounded in responsibility, judgment, and concern for another person.
Hospitals have long had to contend with ethical dilemmas like the ones you described. But are their processes for identifying and handling potential issues sufficient for the AI era? If not, what needs to change?
Many of the tools to address ethical concerns already exist. The greater challenge in the AI era is recognizing when an ethical issue is emerging in the first place. That starts with having a clear strategy and framework for how AI is deployed and evaluated, because implementing technology simply for the sake of implementation can make it difficult to recognize when something is going wrong.
Historically, healthcare has relied heavily on frontline staff to act as the “canary in the coal mine” by identifying concerns, inconsistencies, or unintended harms. If clinicians become overly disconnected from decision-making and defer uncritically to the amoral processes of AI, we risk losing that important signal. That is why thoughtful governance, ongoing human engagement, and training clinicians how to appropriately question and assess AI outputs are so important.
