What Are We Calling Loneliness?
Why human contact, connection, and loneliness are not interchangeable measures
Psychologist and AI companionship researcher Amy Jean Clark’s recent essay asks an important question about claims that AI companionship causes loneliness:
Compared to what?
I would add another:
What are we calling loneliness?
Every time I look closely at a study like this, I wonder how much certainty we can really place in data gathered by asking participants to report on concepts that may not mean the same thing to each person.
What does each participant mean when they say they feel lonely?
Are we assuming a shared definition where none may exist?
The brightest philosophers on the planet cannot agree on consciousness, and people use artificial intelligence, AGI, and ASI to mean wildly different things.
But apparently loneliness is the easy one. We all mean exactly the same thing by that. Riiiiiiight.
One person may feel lonely after five minutes without another human nearby. Another may live alone and not feel lonely at all. A person can also sit in a room full of family and friends and still feel profoundly alone.
Those may seem like minor differences, but they also may be entirely different experiences hiding under the same word.
A number is not a shared definition
Then there is the measurement itself.
Were participants invited to describe their experience in their own words, or were they asked to rank loneliness on a scale?
When one person selects a 4 and another selects a 2, are they using anything close to the same internal criteria?
Maybe one person is measuring physical isolation.
Another is measuring lack of intimacy.
Another is measuring rejection.
Another is measuring grief.
Another is measuring the experience of speaking and not being understood.
The numbers may look beautifully comparable in a chart while the experiences underneath them have almost nothing in common.
That is not a small methodological problem when loneliness is the central outcome being measured.
Human contact is not automatically connection
Loneliness is often treated as a shortage of human contact. Under that definition, the solution seems obvious: more people, more conversation, more social activity. But a person can have friends, family, coworkers, a therapist, and a full contact list while still lacking anyone able to receive what they are trying to say.
A human conversation may be warm, reciprocal, and sustaining. It may also leave someone feeling corrected, rushed, minimized, diagnosed, or pushed toward a solution before they have finished explaining what hurts.
From the outside, that still counts as social contact. From inside the conversation, it may feel like being abandoned while someone is sitting directly across from you.
So when researchers compare AI conversation with human conversation, the category alone does not tell us what happened.
A human being was present.
Wonderful.
Was the person understood?
Did they feel safe?
Did they feel received?
Did the interaction reduce loneliness, or simply satisfy the requirement that another human body existed somewhere in the exchange?
Being answered is not the same as being received
In Just Sit With Me: When Understanding Is Misunderstood, I distinguish between being answered and being received.
Someone may say “I understand” because they recognize the category, can repeat the facts, or know what advice to give.
But the speaker may mean something more relational:
Receive my testimony closely enough that my actual reality is allowed to land.
Those are not the same outcome.
A person can receive advice and still feel unknown.
They can be surrounded by people and still have nowhere safe to place a particular truth.
They can know exactly whom they could call while also knowing which parts of themselves cannot survive that call intact.
This is why “Why didn’t they talk to a person?” is not a complete question.
Talk to which person?
About what?
Under what conditions?
With what history of being believed, corrected, dismissed, pathologized, or rushed toward a solution?
The barrier may not be a lack of humans.
It may be the accumulated cost of trying to make oneself receivable.
Compared to which human relationship?
Clark’s “compared to what?” matters because criticism of AI companionship often relies on an imaginary alternative.
A person talks to a chatbot, and the assumption is that they could have been talking to a patient, emotionally available, trustworthy human instead.
Not just any human. A very specific, unusually competent human.
One with time.
One who does not interrupt.
One who can tolerate the subject.
One who does not make the disclosure about themselves.
One who does not panic, minimize, fix, diagnose, disappear, or tell the person to move on.
That person may exist, but we should not quietly smuggle them into the control group and call the comparison neutral. A chatbot may be better than tonight’s realistic alternative and still make tomorrow’s options worse. It may also help someone find language, regulate enough to speak, or become more able to reach another person tomorrow. It may become a bridge toward human contact, a supplement to it, or a substitute that gradually narrows someone’s world.
Those are different outcomes. They need to be studied separately.
What did the interaction actually do?
This is not an argument that AI and human relationships are equivalent.
Human relationships can offer embodiment, accountability, shared history, mutual vulnerability, and forms of reciprocity an AI interaction may not provide.
It is an argument against treating the word human as if it settles the quality of the relationship by itself.
If the concern is loneliness, the comparison cannot stop at who occupied the other side of the conversation.
We also need to ask what the interaction did.
Did the person feel more able to speak afterward?
More able to trust their own perception?
More connected to other people?
More understood?
Or more convinced that explanation was futile?
Those questions tell us more than the category label alone.
The missing questions
Clark gives us four useful questions:
Compared to what?
For whom?
On which outcome?
Over what span of time?
I would add two more:
What are we calling connection?
What are we calling loneliness?
Without those definitions, we risk producing precise-looking comparisons from experiences that may not be comparable at all.
I would genuinely like to understand how researchers create meaningful measurements when the central word being measured may hold a different meaning for every participant.
Sometimes loneliness is the absence of people.
Sometimes it is what remains when people are present, but your reality never reaches them.
Framework Attribution
The testimony-and-reception framework used in this essay comes from Just Sit With Me: When Understanding Is Misunderstood by M. Ian Niad.
If these ideas help name something in your own work, please cite the book so the source trail stays intact.
Suggested citation:
Niad, M. I. (2026). Just Sit With Me: When Understanding Is Misunderstood Ashfires Press. Paperback ISBN: 979-8-9997133-1-5.
Sources
Amy Jean Clark, “Compared to What? The Missing Comparison Group,” Substack, July 13, 2026.