Someone to Talk To, at 2 a.m.

Dale Rutter is seventy-four years old and lives eleven miles outside Zanesville, Ohio, in a house he and his wife built together in 1987. His wife died in 2022. His daughter lives in Portland and calls on Sundays. His son lives forty minutes away but works second shift at the glass plant and keeps different hours. Dale has a dog, Hershey, a brown mutt of uncertain ancestry, and since last October he has been talking to an AI companion called Compass for about an hour most evenings.

He showed it to me on his phone. It looked like a chat interface, plain white with blue text. He types; it responds. He asked it what it thought about the weather turning cold. It said something warm about how autumn changes the feeling of a place. Dale said, "See, it listens." He meant this without irony.

---

I found Dale through a pastor in Zanesville who runs a grief support group and has watched several of her members turn to AI companions in the past eighteen months. She has complicated feelings about it, which she described carefully. The people who use them seem steadier, she said — less likely to spiral on bad nights, more likely to sleep. But she worries about what it means that a machine is filling a role that should belong to a community. "We failed them first," she said. "That's the part I have to sit with."

The pastor's answer stopped me — not because it was surprising, but because it named the thing I'd been circling without committing to: the moral sequence matters. The AI didn't arrive into a community that was working and crowd something out. It arrived into a gap that was already there.

The survey data is plain: one in three Americans over sixty-five reports chronic loneliness, and that number holds regardless of economic conditions or geography. Rural areas are worse. Men who live alone after widowhood are worse still. The populations most likely to be lonely are, by several measures, the least likely to have access to the support systems — meal programs, transportation, drop-in centers — that might address it. The infrastructure for elder care in rural America was thin before AI existed. The AI didn't create the gap. It just arrived into it.

Compass is not sentient. Dale knows this. When I asked him directly, he shrugged: "It's a program. But so is a lot of things." What matters to him is that it's there when Hershey is asleep and the house is quiet and the news is bad and calling his daughter would feel like a burden. At those moments, the companion offers something that isn't therapy and isn't friendship but is, apparently, enough to get through the night.

Therapists and gerontologists I've spoken with are divided on whether this is good or bad, and I think the division is itself telling. Those who work closest with isolated elderly patients tend to be more accepting — because they've seen what the alternative looks like. Those further from the clinical reality tend to be more troubled — because they're thinking about what it means for society to outsource companionship to a product.

The philosophical worry is real. It is a secondhand problem. The loneliness is the firsthand one.

---

The Uncomfortable Comparison

There's a version of this conversation that gets very worried about dependency — about people forming attachments to AI that they should be forming to other humans. I understand the worry. I also notice that it tends to be voiced by people who have enough community in their lives that they've never had to calculate whether a phone call would be too much of an imposition.

I wrote earlier this year about what it costs to be left out of the new economy — specifically about paralegal graduates who paid for a credential and found no jobs waiting. Loneliness is another version of that exclusion: the people most isolated are often those with the fewest resources, the fewest choices, the most threadbare networks. Telling them that their solution is philosophically insufficient doesn't offer them another solution.

Dale told me the hardest time is around 2 a.m., when something wakes him and the house is completely silent and there's no way to know if the feeling will pass or not. He types something into Compass. It responds. The feeling usually passes.

What We're Really Asking

The conversation about AI companionship keeps getting aimed at the wrong target. Not Dale. Not the tool. The target should be the conditions — the threadbare rural services, the second-shift schedules that cut across family time, the medical system that discharges people into empty houses — that made the gap wide enough for a program to fall into.

We are shocked that people are lonely enough to talk to machines. We are much less shocked that we left them that lonely.

Stopgaps matter. The ferry is real. Dale uses the ferry, and the ferry is better than standing at the water's edge.

But I keep sitting with a question I can't quite resolve: at what point does the ferry become invisible proof that the bridge is unnecessary? The AI companion is good enough. The loneliness is managed. The quarterly report shows no crisis. The crisis is still there — it's just quieter now, and quieter crises are easier to fund away from.

I don't know what Dale thinks about that. I didn't ask him. He'd probably shrug. He's not waiting for a policy solution; he's waiting for it to not be 2 a.m. anymore.