The accelerating advances in artificial intelligence (AI), coupled with ever-increasing pressures for administrative efficiency in health care, are disconnecting doctors from their patients to the point where physicians know less and less about their patients’ lives and needs, sociologist Allison Pugh, PhD, said in an address at Learn Serve Lead 2025: The AAMC Annual Meeting, on Sunday, Nov. 2.
Pugh, a professor of sociology at Johns Hopkins University, in Baltimore, warned about the erosion in medical care of “connective labor,” which she defined as “the forging of an emotional understanding of another person to create valuable outcomes.” That understanding requires empathetic listening, deep interactivity, and “reflecting” back to the other person an understanding of their thoughts and feelings, she explained in a session entitled Seeing the Other: Human Value in the Age of AI.
But greater reliance on AI for interactions with patients, along with growing administrative tasks, leave doctors with less bandwidth to engage in such deep conversations and reflections, Pugh said in her address and subsequent discussion with moderator Susmita Pati, MD, MPH, chair of the Department of Pediatrics at the University of Florida College of Medicine – Jacksonville.
“The stakes are high,” said Pugh, author of a recent book, The Last Human Job: The Work of Connecting in a Disconnected World. “Seeing each other affects not only the individuals involved, but our social world. As people increasingly feel unseen, we all face what we might call a depersonalization crisis.”
Pugh urged doctors to beware of how AI is “being sold by the hype machine that is coming out of Silicon Valley,” which claims that AI will improve health care because it is available to patients around-the-clock; is more empathetic than humans; and will give doctors more time to spend in meaningful conversations with patients. She said those promises will go largely unfulfilled. For example, she predicted that health systems will fill the extra time that doctors have by increasing their patient loads.
Pugh urged doctors to be aware of these dangers when they use AI, to not let it dissolve their patient relationships, and to try to work for companies “that really take this [risk] seriously.”
In an earlier interview, Pugh spoke with AAMCNews about her book and how physicians can maintain human connection in an increasingly disconnected world.
This interview has been edited for brevity and clarity.
What is connective labor in health care, and why is it important?
Connective labor is the practice of seeing the other [person], so that the other feels seen. It’s important for health care in a couple of important ways.
First, diagnostics rely on getting good information. If your patient is not telling you the truth because they don’t trust you, or they don’t think you see them, or they feel misrecognized in some way, then you’re not going to have what you need to make a good diagnosis.
Also, a lot of health care relies on patient compliance, which is all about relationships. Primary care physicians know this well.
Finally, it’s really important for workers. I heard time and again that it [human connection] is what makes the work really meaningful and sustaining for them.
You write, “Connective labor is a valuable human practice under siege by systems that to some degree enable but often impede it.” What are those systems and how do they enable or impede collective labor?
In medicine, the systems are those that are about endlessly counting what people are doing. Too often the people who are charged with connecting are also charged with collecting the data on those relationships, on those interactions [with patients]. Many doctors that I interviewed were proud of how they learned how to look you in the eye and type [notes into an electronic records system]. That’s an impediment. The relentless requirement to account for everything you’re doing and what you’re not doing.
You see scripts — the sequences of questions that doctors are expected to ask — and standardization — that is, certain protocols and procedures — as contributing to this dehumanizing process. How?
The [potential] harm is the dehumanizing of the person. Because the engine behind the scripts is the drive for efficiency. If deadlines and efficiency pressures drive the system, then the temptation is great to minimize or dehumanize the other along the way.
There is a balance. Too often we are off balance. The problem with scripts and standardization is that they lead to degradation of connective labor. And as soon as you go down that slippery slope, you get people saying, “AI can do better than that.” It leads to the automation of these vital human tasks. It’s contributing to mechanization.
You write: “The biggest impediment to warm, empathetic, connective labor is sheer overload. Sometimes workers have too many clients with too much need while they face too many tasks.”
AI advocates contend that the technology lessens that overload. Like apps that record the discussion in a physician-patient visit so the doctor doesn’t have to type while they are listening. Or chatbots with my health care system that give me answers about certain conditions. The idea is that we free up doctors from the laptop during visits and from online messages so that they can connect in a more focused, personal way when we meet. But you write that “the free-up language seems strikingly optimistic.”
The first problem is that these people are very optimistic about modern capitalism. But if a machine can do your job, then you’re going to lose it. In the case of physicians, they’re going to get piled on with different work. There has never been a case in which our medical system has eased up on what we demand of physicians.
Point number two: These free-up claims are unrealistic about what AI can do well. I’ve spoken to physicians who have been asked to rely on AI scribes. I’m thinking of a physician [treating] diabetes. He said [regarding a patient meeting] AI ignored all the connective labor and started taking notes after some 20 minutes had gone by. The connective labor [that AI ignored] was along the lines of, “Are you still living with your sister? Is she bringing you to the grocery store?” That’s relevant information for a physician. It [the AI tool] was ignoring what it didn’t think was medicine, but what is actually medicine for some physicians.
When I go to see my doctor now, he does spend less time typing and more time looking at me because an AI tool is listening and transcribing the conversation into his notes.
Which is a positive!
So can AI contribute to connective labor?
I want to say yes. I’m not immune to this transformation. I’ve heard some technologists advocate [for AI] as a way to practice hard conversations. That’s not a terrible idea, and it is probably good for students or trainees to practice how to say something that’s difficult.
I appreciate that you write about grocery stores and self-checkout — about how checkout clerks used to talk with you about what you need.
Grocery clerks, a hundred years ago, used to do connective labor. They would talk to you, they knew who you were buying for.
I love going to self-checkout at grocery stores. You made me feel bad about that.
Good! I wrote this book to highlight what road we’re on and what the choices we’re making today are going to lead to if we’re not careful. I reject that [the choices] are inevitable.
You call for “a new awareness of connective labor and a social movement to protect it.” What does such a social movement look like in medicine?
Administrators have the power to make change happen. That ranges from paying close attention to time, space, and money, but also to the norms and values that you are propagating.
The organizations that I found were doing it right were careful about giving people who do connective labor chances to talk to others about it. That’s hard. It’s another meeting. It’s another thing that doctors have to give time to.
Organizations that do it well, they take attention seriously. It means limiting how much a worker has to attend to, not overloading them with appointments or tasks. Treating that attention as if it’s precious.
I’m not saying they have to eschew all technology or systems. Just subordinate them to the real priority of connection, not the priority of efficiency. We’re out of whack.
If I’m a health care provider, I’ve still got metrics to meet. How do I measure the value of what you’ve just asked me to do?
The domination of metrics is a symptom of our pathology. We are letting the metrics and the insistence upon data analysis drive the bus. The human interactions should drive the bus. We should fit the other stuff in as it helps that bus go in the right direction.
