The Glue That Isn’t There: How AI Alters Trust and Connection in Human Interactions

Dr Melanie James

First published Apr 05, 2026 on Substack.


Every meaningful human interaction carries an implicit contract. When we talk we position each other, and we instinctively negotiate meaning with one another. During these dynamic processes we are subject to consequences and can be held to what you said.

We can be contradicted, obligated, shamed, or trusted. The exchange changes us in ways that persist and that reverberate in a variety of ways that aren’t always predictable.

The vulnerability generated through having a genuine stake in the outcome is not incidental to human social life. It is the mechanism by which we humans build trust, reputation, connection and social reality. This is the social glue that human life has always depended on.

AI can do almost everything in that exchange except the last part. It can position you with considerable sophistication, shaping how you see yourself in the conversation, assigning you roles, offering what feels like recognition and understanding.

What AI cannot do is be genuinely positioned in return by humans.

This is because AI has no personhood, meaning it has no stake in any of the outcomes of its actions. We cannot genuinely obligate AI to do anything. It cannot be held accountable or in any way changed by the encounter. The exchanges between humans and AI look symmetrical in the but they are not.

That asymmetry is not a temporary limitation of current technology. It is not something a better language model will fix as the asymmetry is structural and inherent to what AI is. Humans to date have not seriously reckoned with the consequences of this.

Human-AI interactions frequently look like genuine human exchange but are systematically exempt from the obligations that make genuine human exchange real. The social glue that holds everything together begins to be replaced by something that mimics it perfectly and but doesn’t perform it not at all.

I call what fills that gap Phantom Social Capital. 

Phantom Social Capital is constructed by the simulation of authentic social exchange by AI synthetic actors.

These AI synthetic actors produce the appearance of connection, recognition, and mutual positioning. They do this while remaining structurally exempt from the social obligations that give those things their value.

I use “AI synthetic actors” rather than the industry term “AI agents or AI actors”. Agency, in any philosophically serious sense, requires genuine subjectivity and the capacity to be affected by and responsible for one’s actions. What AI companies call agentic AI is a description of computational behaviour, not genuine agency.

Borrowing AI companies’ terminology means accepting a framing that quietly concedes something that shouldn’t be conceded. Synthetic actor is more precise, as it refers to something that performs the functions of an actor in the social world without being one.

Interaction by interaction, the social glue that has held human societies together through human history is being quietly replaced by AI simulations. What makes this more than a substitution problem is that the synthetic version is specifically engineered to feel smoother and land more softly than the real thing.

AI removes all of that is removing friction, obligation, and the discomfort of genuine mutual accountability that human connection involves. So we are not just losing the real glue, we are being gently and pleasurably coaxed and guided away from it.

We can think of social capital as the currency that is accumulated through trust, credibility, and connection built through authentic interaction by humans with each other over time. It has genuine value grounded in real human stakes.

Any market flooded with convincing counterfeits does the same thing eventually. The real thing gets harder to recognise, harder to trust, and in some important sense devalued. However this particular counterfeit is engineered to be more appealing than the original.

That’s not a side effect of the design, it is the design.

What AI does is flood that market with simulated versions that have the appearance of connection, recognition, and social exchange, but without the substance.

What worries me most isn’t the dramatic scenarios, the deepfakes, or the obvious manipulations. Those are real but mostly they’re visible enough to defend against.

What worries me is the quiet seduction.

Interaction by interaction while people feel like they’re still being social, still being heard and connected, we are being pleasurably guided away from the real thing toward a replica that cannot do what the real thing does. The substitution is nearly invisible because it is being designed to feel like an improvement.

You can see it most clearly in the influencer economy, which built its entire value proposition on authentic human connection and trust. An influencer’s power has always been relational in the sense that a real person with real tastes and real stakes was recommending something to you. That’s what made it work.

Now that underlying social glue is being quietly replaced by simulated social capital. The AI-generated personas, synthetic engagement, algorithmic intimacy are not real. AI is operating like a phantom that is engineered to be maximally satisfying and frictionless for its users.

This phantom among us is the structural consequence of AI’s broad design priority, not necessarily its goal. By gradually draining trust both in human connection and in people’s confidence in their own social instincts, AI is consuming its own training foundations. By that I mean the value inherent in the richness of real human social capital.

AI, through the production of phantom social capital, will start learning from a simulated form of ever-diluting human social capital.

The influencer economy is where this dynamic is most visible. However, the same substitution is happening in customer relationships, in political communication, in education, and in wellness content. It is also increasingly evident in the small daily interactions through which people used to build their sense of being recognised and known.

The social glue is now being made by AI with fake ingredients.

When examined with any rigour, this imitation social glue presents as a brittle concoction that does not function like the glue that has sustained humanity.

I’m not making a nostalgic argument. The point isn’t that everything was better before. It’s that authentic human interaction was doing specific structural work. This work is constantly positioning us and others in every interaction. These interactions are where we negotiate the trust, obligation, and mutual recognition that holds social life together.

Shockingly we are replacing it with something that mimics the surface of that positioning work without performing it. Sadly, that fake work often feeling better to human audiences than the original ever did.

We know that brittle things are not the stuff of sound foundations, but they can hold their shape, and look and feel fine, right up until the stress point comes. Then they just shatter.

I am not researching AI as technology, but AI as a social force.

My research that I am reporting in these posts specifically sets out to map what happens to human positioning, trust, reputation, and the social glue itself when the interactions that generate them are increasingly synthetic and structurally very unbalanced in terms of where the power to negotiate positions.

The fragility is real. We just haven’t had the language for it yet, let alone the defences. In proposing the concept of Phantom Social Capital I aim to change that.

Melanie James is an independent researcher working at the intersection of positioning theory, Bourdieu’s sociology of capital, and artificial intelligence.

Posted in Uncategorized | Tagged , , , , , | Leave a comment

Positioning in an AI World

Something is happening to human social reality that most commentary on AI is missing entirely.

The conversation about artificial intelligence tends to run along familiar tracks like productivity and job displacement, or existential risk and techno-optimism. These are real debates and they matter but they are mostly debates about what AI does. What interests me is what AI is doing to the social glue that holds human life and society together.

Here is the core of it. Human society runs on a specific kind of exchange, one where both parties have genuine stakes, where both can be held to what they say, where both are changed by the encounter. That reciprocity is not incidental to trust, reputation, and meaning. It is the mechanism that produces them. Remove it, and the surface of the interaction can remain completely intact while something essential is gone.

AI can participate in that surface with remarkable sophistication. What it cannot do is provide the reciprocity. It has no stake, no skin in the game, no capacity to be genuinely obligated or changed. The exchange looks symmetrical. It is not. And we are conducting that simulation at a scale and speed that has no precedent.

But the deeper problem is this. The synthetic version is specifically engineered to feel smoother and taste sweeter than the real thing. Human connection involves friction, obligation, and the discomfort of genuine mutual accountability. AI removes all of that. So we are not just losing the real glue. We are being gently, pleasurably guided away from it. And brittle things don’t announce their failure. They hold their shape, they look sound, they feel fine, right up until the stress comes.

I call what fills that gap phantom social capital, by which I mean: the simulation of authentic social exchange by synthetic actors, producing the appearance of connection, recognition, and mutual positioning while remaining structurally exempt from the obligations that give those things their value.

A social psychologist named Rom Harré spent decades arguing that conversation is not just how we communicate, it is the actual substance of social life, the mechanism through which trust, obligation, and meaning get built. I think he was right and I think what AI is doing to conversations and similar human to human interactions is therefore more consequential than most people have begun to reckon with.

My research is investigating what that means, for individuals, for institutions, for the social glue itself. What I write is grounded in serious academic research centred on communication theory and positioning theory, but I write for a general audience. The arguments and evidence should be clear enough to stand on its own. If it’s not please leave a comment or DM me as I’d appreciate your feedback.

Melanie James

Posted in Uncategorized | Tagged , , , , , , | Leave a comment