One of the things that stayed with me from Paul Kingsnorth is not the technology argument at all. It is simpler and harder.
Humans are built for community, not individuality.
We used to know our neighbors. We had friends we could call at two in the morning. We showed up for each other at work and at home.
Now, broadly, we mind our own business, type unkind things at strangers online, and stream shows until two in the morning by ourselves.
That is not a technology essay. It is a description of what happened while we were being connected.
The promise and the result
Social platforms were sold as connection. What they delivered was contact. Enormous quantities of contact, with a very low proportion of the thing that actually feeds people.
I don't think anyone at those companies set out to isolate a generation. That is what makes it worth studying. The outcome wasn't chosen. It emerged from a lot of individually sensible optimizations, each one improving a metric, none of them measuring whether anyone was less lonely.
Nobody was tracking two in the morning friendships. So nobody noticed them going.
Why automation faces the same test
Here is where this gets uncomfortable for me, because it is my industry.
Automation, done carelessly, removes human contact from processes. That is often exactly the point and often fine. Nobody needs a human being to time-stamp an invoice.
But the same move applied slightly wrong takes the human out of the moments that were carrying the relationship.
The check-in call that became an automated email. The person who used to notice you sounded off and now is a form. The receptionist who knew your kid's name, replaced by a routing menu that's measurably more efficient and has never known anything about anyone.
Every one of those decisions saves money and each looks defensible in isolation. Together they hollow out the thing the business was actually made of, and it happens too slowly for anyone to point at the day it broke.
The test I try to apply
When I'm scoping automation, the question I try to hold isn't can this be automated. It is: does this return people to each other, or remove them from each other?
Automating invoice entry returns people to each other. The owner who was doing data entry at nine at night is now available to their family, or to their team, or to the customer who needed a real conversation.
Automating the customer conversation itself removes people from each other. It might still be the right call for some businesses. But it should be a decision someone made deliberately, knowing what's being traded, not the accidental result of optimizing a queue.
The tell is usually who benefits. If the efficiency accrues to the company and the cost lands on the person with less power to object, that is worth pausing on. Not automatically wrong. Worth pausing on.
What technology is worth
A tool is worth what it gives back to the people who use it.
Not what it produces. Not what it saves. What it gives back, and to whom.
By that measure a lot of what we have built in the last two decades has a worse return than its adoption numbers suggest. We got extraordinary capability and, if Kingsnorth and the loneliness data are pointing at anything real, considerably less of the thing that makes a life good.
I do not think that means we should build less. I think it means we should be much clearer about what we are aiming at, because the tools don't aim themselves and the metrics we default to aren't measuring it.
The version I actually believe
The best possible outcome for this technology is unglamorous.
Not that it does astonishing things. That it quietly takes the machine work off people so they have more of themselves left for each other. Fewer evenings lost to paperwork. Fewer relationships conducted in the gaps between administrative tasks. More capacity for the parts of work and life that only a person can do.
That is available. It is also not automatic, and it will not happen by optimizing throughput, because throughput has never once measured whether anyone had someone to call at two in the morning.
Whatever you're building or buying: does it return people to each other, or take them further apart?
That is a harder question than whether it works. It is also the one that will matter in ten years.