Sherry Turkle's 'Artificial Intimacy' & NEA as the Antidote

Sherry Turkle has spent her career watching what happens when we let technology into the most human spaces. Her verdict is not sentimental technophobia — it is a carefully evidenced argument that simulation corrodes the real. If you work in chaplaincy and you're thinking about AI, you owe it to your patients to take her seriously.

And then you owe it to your patients to understand why NEA is the opposite of what she's warning against.

I. Turkle's Warning

Sherry Turkle is a clinical psychologist and professor at MIT's Program in Science, Technology, and Society. Her 2015 book Reclaiming Conversation documented something quietly alarming: the more we offload emotional interaction to screens, the less capacity we have for the real thing. In subsequent work under the heading "artificial intimacy," she has pressed this argument further into healthcare — into the precise territory where chaplains live.

Her concern is not that AI is clumsy or unreliable. Her concern is more unsettling: that AI is too comfortable. It never has a bad day. It never needs anything in return. It doesn't get tired of your grief or run out of patience with your questions. When we offer AI companions to the elderly, AI therapy chatbots to the depressed, AI consolation to the dying, we are not providing imperfect-but-helpful substitutes for human care. We are training people — and institutions — to prefer the manageable simulation over the demanding, reciprocal, irreducibly human reality.

The risk compounds at the institutional level. A hospital that deploys an empathy bot for distressed patients isn't just providing an option — it's making a structural choice that human presence is optional, substitutable, scalable. The bot costs less. It doesn't call in sick. It never needs supervision. Over time, under financial pressure, the human gets cut and the bot remains. What we lose in that transaction is not efficiency. It's something we can't recover.

Turkle's sharpest observation is this: we don't just harm the care-recipient when we substitute AI for human presence. We harm the care-giver. The chaplain, nurse, or social worker who is replaced by a simulation doesn't simply lose a job — the whole institution loses the capacity for genuine clinical relationship. Skills atrophy when they're not practiced. And when we habituate to simulation, the real begins to feel effortful, risky, uncomfortable by comparison. The simulation wins not through quality but through convenience.

II. Why This Critique Matters for Chaplaincy

Chaplaincy sits at a peculiar intersection. It is a clinical discipline with rigorous educational and credentialing standards — board-certified chaplains complete a master's-level degree, clinical pastoral education residencies, and peer examination. And yet its clinical value is inseparable from something that cannot be operationalized in a protocol: the experience of being genuinely present with another person at the margins of their life.

A chaplain cannot simply ask the right questions in the right order and thereby provide spiritual care. The patient or family member who is facing a terminal diagnosis, who is asking whether their suffering means something, who is estranged from their faith and doesn't know what to do with that estrangement — they are not asking for information retrieval. They are asking to not be alone with the weight of it. The care is constituted by the presence of another human being who genuinely receives what the patient is saying and is not destroyed by it. That is irreducible. It cannot be approximated.

This is exactly why Turkle's warning applies in its fullest form if AI enters chaplaincy as a substitute for human presence. An AI chaplain — a system that converses with patients about their spiritual distress, that offers scripted empathy, that asks the questions a chaplain would ask — does not provide a diminished version of what a chaplain provides. It provides something categorically different that happens to use the same vocabulary. And it teaches patients, families, and institutions that the vocabulary is what matters, not the presence behind it.

If you are a chaplain reading this, you already know all of this. The question worth sitting with is not whether AI can simulate your work — it obviously can simulate the surface features of it. The question is whether the institutions you work in will be able to hold that line under financial pressure. Turkle's warning is not primarily addressed to you. It's addressed to hospital administrators, payers, and policymakers who might look at an AI chaplain and see a cost-effective alternative to a difficult-to-bill human service.

III. The Distinction Turkle Misses

Turkle's critique is powerful and largely correct. But it contains a hidden assumption that matters enormously: it assumes AI is deployed at the point of human connection — in the moment of care, substituting for or mediating the human encounter.

NEA operates entirely outside that moment.

The NEA framework assists chaplains after the encounter ends. When the chaplain leaves the room, when the patient is resting or has been called away for a procedure, when the chaplain is back at their workstation processing what happened — that is when NEA enters. The AI never enters the room. It processes documentation, reflects on the quality of the encounter, identifies patterns across visits, and helps the chaplain understand their own development. The patient never interacts with it. The care relationship is never mediated by it.

This is a categorically different use of AI. Turkle distinguishes between AI that is onstage — present in the human encounter, performing relationship — and tools that are backstage, supporting the human who then performs the relationship themselves. She focuses almost entirely on the onstage case. NEA is backstage in the most rigorous possible sense: the encounter is over before NEA exists.

The analogy is not perfect, but consider the musician who records and reviews their own performances to improve. The recording technology is not present in the live concert. It does not mediate the relationship between performer and audience. It serves the performer's development after the performance, so the next performance is better. NEA works this way: documentation as reflection, reflection as growth, growth as better presence next time.

IV. Documentation as the Enemy of Presence

Here is the irony that most AI-in-healthcare critics miss: the greatest structural threat to chaplain presence is not technology. It is the existing administrative burden — and technology designed correctly can address it.

Consider what chaplains currently navigate. After a six-hour shift that may include a trauma activation, two end-of-life family meetings, a suicide-attempt follow-up, and a rapid-response consult, a chaplain returns to their workstation to chart every encounter in a system designed for nursing workflow. The notes must satisfy billing compliance, Joint Commission documentation standards, and department-specific requirements that vary by institution. Two hours of charting is not unusual. On particularly heavy days, more.

This creates what presence researchers call the divided mind. The chaplain who is still at the bedside but already composing the note in their head — already finding the words that will satisfy the documentation requirement — is not fully present. The administrative layer intrudes into the care layer. This is not a chaplain failing to be present; it is a structural condition that makes full presence harder to maintain. The note has to get written, and the mind that writes it begins writing during the encounter itself.

When NEA handles the documentation layer — when a chaplain can complete their charting in minutes rather than hours, and can do so with their reflective attention rather than their administrative attention — the effect is not efficiency. The effect is presence. The chaplain who is not dreading two hours of charting at the end of a shift has more to give in the final encounter of that shift. The chaplain whose documentation is completed quickly has more cognitive space to actually reflect on what happened, rather than just recording that something happened.

This is the reversal of the Turkle critique. The question is not whether AI threatens presence. The question is whether, designed correctly, AI can protect and even increase it.

V. What NEA Protects

The NEA framework — Noticing, Engaging, Accompanying — was not designed as a documentation taxonomy. It was designed as a theory of presence. Each dimension describes a quality of attentiveness that is either present or absent in a chaplaincy encounter, and the absence of any one of them represents a real reduction in the care that was offered.

Noticing is the quality of receptivity the chaplain brings to the encounter: seeing what is actually in the room, reading what the patient is communicating non-verbally, registering the emotional content of a silence, perceiving who is present in the family system and who is absent. Engaging is the quality of contact — whether the chaplain's responses land, whether they meet the patient where the patient actually is rather than where a protocol would place them. Accompanying is the quality of sustained presence through difficulty — whether the chaplain remained with the patient through the hard parts rather than redirecting away from them.

When NEA produces a reflective analysis of an encounter, it is not producing a performance review. It is producing a mirror. The private mentor feature doesn't ask, "Did you complete the required elements?" It asks: "Where were you most present in this encounter? Where did your attention drift? What was the patient trying to tell you that you may not have fully received?" These are questions the best clinical supervisors ask. They are questions of formation, not compliance.

This inverts the Turkle critique entirely. Turkle worries that AI trains us away from authentic relationship by making simulation comfortable. NEA is designed to train chaplains toward authentic relationship by making its quality visible and reflectable. The measure of whether NEA is working is not whether documentation is faster — it's whether chaplains become more present over time because they have a rigorous tool for understanding their own quality of presence.

VI. The Design Ethic We're Committed To

There are AI products in development — some from well-funded companies, some from academic research groups — that would deploy conversational AI in direct patient interaction for spiritual care. Chatbots that ask "how are you feeling about your diagnosis?" Automated check-ins for patients in long-term care. AI companions for dying patients who are alone at night. We understand the impulse behind these products. Chaplain coverage is thin. Nights and weekends are often uncovered. The suffering is real and continuous, and there are not enough humans to meet it.

We are not building those products. We will not build those products.

Our design constraint is absolute: AI must never enter the patient encounter. It processes after, reflects after, and learns after. If someone asks us to build an AI chaplain that interacts with patients, the answer is no — not because we can't, but because doing so would be a betrayal of the thing we exist to protect. The value of a chaplain is constituted by their humanity, their mortal presence, their capacity to genuinely receive suffering because they are themselves susceptible to it. An AI cannot be present in that sense. It can perform presence. Performing presence is not presence.

Knowing the difference is everything. The line between AI that protects human work and AI that replaces it is not always obvious — and in healthcare, getting it wrong has consequences that cannot be undone. Turkle is right that the pressure to cross the line will be institutional, financial, and gradual. A "pilot program" here, a "coverage tool" there, and eventually the human presence that was supposedly being protected has been quietly made optional.

We are building the tool for the chaplain. Not the replacement for the chaplain. The distinction is not a marketing claim — it is a design commitment that has to be renewed with every feature decision we make. When we ask whether a given capability belongs in NEA, the first question is always: does this serve the chaplain's presence with their patient, or does it begin to substitute for it? If the answer is the latter, we don't build it.

Turkle's warning is real. The antidote is not less technology — it's technology that knows its place. NEA is designed to stay in its place: backstage, in service of a human encounter it never touches.