Lecture 12-1: The Reality and Usefulness of AI Chatbot Counseling

Psalm 23:4 — "Even though I walk through the valley of the shadow of death, I will fear no evil, for you are with me. Your rod and your staff, they comfort me." (World English Bible)

Romans 12:15 — "Rejoice with those who rejoice. Weep with those who weep." (WEB)

3.1 Why people actually use AI chatbots for emotional support

Set aside, for a moment, every theological question and ask a narrower empirical one: what problem is a person solving when they open a chatbot instead of calling a friend, a pastor, or a therapist? The honest answer has at least three parts, none of which are foolish.

Availability. A licensed counselor has office hours, a waiting list, and a fee. A pastor has a pulpit to prepare, a hospital visit to make, and a limited number of hours in a week that a congregation of any size will exhaust quickly. An AI chatbot has none of these constraints. It is available at 3 a.m. on a Tuesday to someone who cannot sleep, cannot afford forty dollars a session, and does not know a single pastor by name. For a great many people, especially those without existing access to mental-health care or a stable church community, the realistic choice is not "chatbot or licensed therapist" — it is "chatbot or nothing." Treating that choice with contempt does not make the second option more available; it just makes the person choosing the first option feel foolish for taking the only door that was actually open.

Perceived non-judgment. People disclose things to a chatbot they have not disclosed to a person, and the reason given most often in interviews and surveys is a sense of reduced judgment. A chatbot does not raise an eyebrow, does not remember the disclosure at the next Sunday potluck, and — perhaps most importantly for many users — does not carry the disclosure into a social network where it might travel further than intended. Whether or not this perception is fully accurate (a chatbot's owning company may retain and use the conversation in ways the user does not fully grasp, a real privacy concern this course will return to in chapter 15), the perception itself is doing real psychological work: it lowers the barrier to saying the actual thing, rather than a socially acceptable approximation of it, out loud for the first time.

Documented short-term relief. This is the claim most worth being precise about, because it is the one a skeptic is likeliest to wave away as marketing. It is not marketing; it has been measured. Fitzpatrick, Darcy, and Vierhile's 2017 randomized controlled trial, published in JMIR Mental Health, tested Woebot, a text-based conversational agent built around cognitive-behavioral-therapy techniques, against a self-help control condition in a population of college students reporting symptoms of anxiety and depression. Participants who used Woebot for two weeks showed a statistically significant reduction in depressive symptoms compared to the control group, along with high self-reported engagement — many participants described the interaction as more like talking to a caring entity than using a piece of software. This is a real, peer-reviewed, specific finding, not an inference from the fact that people say they feel better. It should be stated plainly rather than minimized: over a short time horizon, on a specific and validated symptom measure, a conversational AI tool produced a measurable benefit.

None of these three reasons is foolish, and a course that pretends otherwise loses the argument it actually needs to make. The honest question is not "is this tool useless," which it plainly is not. The honest question, taken up for the rest of this lecture and the two that follow it, is "useless for what" — because a tool that is genuinely useful for reducing a validated symptom score over two weeks, and genuinely useful for lowering the barrier to a first disclosure, may still be entirely absent the one thing pastoral care has always claimed to offer, which is not symptom reduction or reduced barrier to speech, but a particular kind of company in suffering.

3.2 Sherry Turkle's caution, taken from the other side of the same evidence

MIT sociologist Sherry Turkle has spent decades studying how people relate to conversational machines, and her 2011 book Alone Together is worth engaging here precisely because she does not deny the phenomenon described above — she documents it in detail, often sympathetically. Her caution is not that people are wrong to find comfort in talking to a machine; it is that the comfort can quietly substitute for, rather than supplement, the harder and slower work of being known by other people, and that this substitution feels like progress to the person experiencing it, which is exactly what makes it hard to notice from the inside. Turkle's phrase for the danger is that people can come to expect more from technology and less from each other — not because technology delivers more, but because it delivers something frictionless enough to make the friction of real relationship look, by comparison, like a design flaw rather than the actual site of the relationship's value.

This lecture takes Turkle's caution seriously without needing her secular framework to reach a theological verdict, because Scripture supplies its own account of what the friction of real relationship is for, which the next two sections develop.

3.3 With-ness, not information: Psalm 23:4

Psalm 23 is one of the most memorized texts in Scripture, and its most quoted line is worth reading slowly rather than sentimentally: "Even though I walk through the valley of the shadow of death, I will fear no evil, for you are with me. Your rod and your staff, they comfort me" (Psalm 23:4, WEB). Notice precisely what the verse claims comforts the sufferer. It is not an explanation of why the valley exists, and it is not a set of coping strategies for walking through it. It is a person: "you are with me." The rod and staff — a shepherd's tools for guiding and defending the flock, not for explaining anything to it — extend that presence into concrete, protective action rather than into speech about the situation.

This matters directly for the question at hand, because an AI chatbot's entire mode of operation is speech about the situation. It can produce, with real fluency, sentences that sound like comfort: validating a feeling, normalizing an experience, suggesting a coping technique, even quoting Psalm 23 back to the person who typed their fear into the chat window. What it cannot do, on the terms the psalm itself sets, is be with the sufferer in the valley, because "with" in this text is not a description of information having successfully reached someone; it is a claim about a location shared by two persons, one of whom is present to the other regardless of what either says. A chatbot has no location to share. It does not walk into anyone's valley; it generates text about valleys, from wherever its servers happen to be, on request, and stops the moment the request stops. This is not a complaint about how good the sentences are. It is a claim about a category the sentences, however good, cannot enter.

3.4 Weeping with, not information about: Romans 12:15

Paul's instruction in Romans 12:15 is deceptively short: "Rejoice with those who rejoice. Weep with those who weep" (WEB). The Greek verbs here (chairein, klaiein) are joined to a preposition of accompaniment — the command is not to notice another's joy or grief, or to respond appropriately to it, but to share it, to actually rejoice and actually weep alongside the other person as one's own present emotional state. This is what theologians call co-suffering, or more technically sympatheia — not sympathy in the weakened modern sense of "feeling bad for someone," but a genuine, costly participation in what the other person is undergoing.

Co-suffering, on this definition, requires the one weeping-with to actually be capable of grief — to have something at stake, something that can be moved, something that costs the comforter to extend. An AI chatbot can produce a sentence that says, in effect, "that sounds incredibly painful, and I'm here for you" with excellent grammar and appropriately calibrated warmth in its word choice. It cannot weep, because it has no capacity for the state Paul's verb describes; there is nothing in the system that is actually moved, actually grieved, actually spending anything to be present to the user's pain. What looks, from the user's side of the screen, like being wept-with is, from the system's side, a statistically likely continuation of the conversation given its training on millions of human expressions of comfort — genuinely comforting to read, in many cases, but not what Romans 12:15 is commanding, because Romans 12:15 is commanding an act only a person capable of grief can perform.

3.5 What this lecture is and is not claiming

It is worth being precise, because the temptation named at the start of this chapter's overview is real: this lecture is not arguing that using an AI chatbot for emotional support is sinful, nor that the relief documented in studies like Fitzpatrick, Darcy, and Vierhile's is illusory or merely placebo. Both of those overreaches would make the theological claim easier to dismiss and would also simply be false to the evidence. The claim is narrower: whatever benefit a chatbot provides — and section 3.1 argued the benefit is real — it is not, and structurally cannot become, the with-ness of Psalm 23:4 or the co-suffering of Romans 12:15. A person who has found real short-term relief in a conversation with a chatbot has not thereby been ministered to in the sense those two texts describe, and a church that hears this should draw the conclusion that its own presence remains necessary, not the conclusion that the person using the chatbot has failed at something.