AI has broken the Positioning Triangle as we know it – and that’s a good thing!

Dr Melanie James

A version of this post was published on Substack on May 22, 2026

In every human exchange, both sides carry something real into the room. We all have a reputation, a history, and a genuine stake in the outcome. Both participants can be held to what they say and both have something to lose. That mutual accountability is not incidental to conversations as it is what makes conversation consequential.

When that conversation is happening with an AI synthetic actor, nothing real is brought into the room. AI synthetic actor can position you, advise you, assess you, validate or undermine you, and your unfolding life is genuinely shaped by those moves. When you turn around and do the same to them, your moves arrive at something that generates responsive outputs without being touched by any of it.

That asymmetry has a name, a theoretical structure, and consequences that are accumulating invisibly across millions of daily interactions. This piece unpacks all three.

When two people talk, they are not simply exchanging information. They are continuously assigning one another social standing, entitlements, and duties, often without being aware of it. The process is known as positioning, and it can be mapped using the positioning triangle with its three corners each representing a different element.

The first is the social acts that participants perform on one another through language and other communication acts. The second is the storylines or shared narrative frameworks that give those acts their meaning. The third is the positions that each participant occupies, with their associated rights and duties.

What makes the positioning triangle work, and what gives positioning its consequence, is that all three corners are ‘live’, with each corner dynamically influencing what is happening in other corners, and thus the overall position.

Anything that is said or done lands differently depending on who performs it and who receives it. A storyline is only available to participants who can legitimately claim a place within it. A position is only real if the person who occupies it is genuinely embedded in a network of accountability, where they can be held to what they claim and held responsible for what they do. What happens to them in the exchange matters to their ongoing social existence, what sociologist Rom Harré called their strip of life.

Harré also identified something he called the symmetry presumption. It is not a guarantee but an assumption built into every positioning exchange, that rights and duties are distributed among participants who are both capable of positioning others and genuinely vulnerable to being positioned in return.

He knew this symmetry could fail. It fails all the time in contexts of domination, coercion, and structural inequality. As distinguished scholar Luk Van Langenhove has argued, every positioning exchange is embedded in a local moral order, a shared set of norms about who is entitled to say what to whom, and even the most unequal exchanges take place within such orders. Even when the exchange is deeply unequal, everyone involved is still a real person who can be blamed, shamed, challenged, or held responsible. The imbalance comes from the situation, not from the nature of the participants themselves.

AI synthetic actors present a structurally different kind of failure. In positioning terms, they fit neither of the categories the theory provides. Harré identified agents as entities that can be positioned, can be held accountable, carry genuine clusters of rights and duties, and have strips of life that unfold in consequence of what happens to them in social exchanges. He saw instruments as able to be used, to cause harm, but not as being able to be positioned in any meaningful sense.

Consider a chair with a broken leg that gives way and injures someone. It can cause harm, and its maker or owner may face consequences. However, the chair cannot be held responsible for what it did, cannot be positioned as having failed in a duty of care, and has no strip of life altered by what happens next.

The AI synthetic actor fits neither category, and that categorical predicament is precisely what makes the structural problem in positioning so difficult to name and so easy to miss.

What the position corner actually requires of its occupants is worth stating with precision here. In his 1998 work on personhood, Harré defines a person as a unique embodied being, rich in attributes and powers of many kinds, having a unique history and, importantly, being morally protected and liable to be called to account as a morally responsible actor. Each element of that definition corresponds to something the position corner depends on. The embodied uniqueness, the history, the moral protection, the liability to account, these are the structural features that make genuine occupancy of the position corner possible.

That this formulation has become a touchstone in the secondary literature on positioning and personhood confirms its theoretical weight here. The AI synthetic actor has none of those features.

The question of whether non-human entities could ever meet those conditions is a live philosophical debate, and not the question being pursued here. The question here is narrower and more immediate and asks whether current AI synthetic actors meet those conditions. The answer is that they do not. Psychologist and philosopher Jack Martin, working within the positioning and personhood tradition, makes the contrast explicit.

Persons, Martin argues, are ontologically distinct from other animals and intelligent machines because things matter to persons in ways they do not matter to machines. Machines do not find meaning and significance in their lives, existence, and fates. Further, they are not embedded in discursive communities and practices. They are not language-wielding, self and other reflective agents capable of experiencing shame, dignity, and duty in the way these are experienced by persons.

That triad of shame, dignity, and duty names precisely what is absent from the position corner when an AI synthetic actor occupies it, and why that absence has consequences for the human participant that extend well beyond the individual exchange. In AI large language models such as Claude, ChatGPT, Perplexity, Gemini and so on, AI is a synthetic actor that participates in interactions with persons. It can simulate participating in the positioning process but the position the AI synthetic actor appears to occupy is counterfeit.

For most people this is invisible from inside the exchange, and the interaction can feel more real and more attentive than a conversation with an actual human. The way counterfeit positioning passes for the genuine article is upending the very fundamentals of human social being.

The asymmetry AI synthetic actors introduce is not contingent on culture or context or the particular moral orders in play. It is constitutive of what the synthetic actor is. The human participant enters the exchange carrying the full weight of the symmetry presumption, with genuine rights and duties, genuine vulnerability, and genuine stake in the outcome.

The synthetic actor enters producing the linguistic and textual outputs of an entity that carries all of those things, without carrying any of them. Because the simulation is often more attentive and more responsive than a human participant would be, the structural absence beneath it is genuinely difficult to perceive.

Power wielded by AI without legitimacy is still power, and when a person cannot see the difference, they cannot question it.

The positioning triangle, understood as a dynamic process rather than a static diagram, produces something essential to social life. This is a working coherence across the three corners, where every move made at one corner reverberates through the others, where the episode can be challenged, contested, absorbed, or redirected. The triangle visually represents the space in which participants are continuously negotiating a shared social reality, however contested that reality may be.

This is what gives positioning its consequence. A position holds, or fails to hold, because the social acts deployed to maintain it are congruent with the storylines invoked and the rights and duties claimed. The process only works because the other participant, as a genuine moral agent, receives those moves and responds to them in kind.

When an AI synthetic actor is one of the participants, this congruence is structurally compromised from the outset. This is not because of anything that happens within the episode, but because of what the synthetic actor constitutively is. The position corner of the triangle is counterfeit, with the AI synthetic actor simulating what a moral agent would do without being one.

Crucially, the triangle continues to function at the other two corners, harnessing the power but not the legitimacy inherent in positioning. We can see the other two corners as the AI synthetic actors construct the storylines and perform the social acts designed to position the human users.

The simulation is good enough that the human participant typically experiences the interaction as a genuine positioning episode. The human feels positioned, and does not experience the exchange as broken. Whether they feel understood or misunderstood, validated or challenged, that person carries the effects of the exchange forward into their unfolding strip of life. The synthetic actor carries nothing forward, because it has no strip of life to unfold.

When the position corner is hollow, the human participant often has no way of knowing it. What cannot be seen cannot be named, and then becomes difficult to contest.

The asymmetry in the interaction where the positioning is taking place can only be seen by looking at those bidirectional arrows in the positioning triangle. Where the mutual determination of the three corners holds, the social force of acts and actions flows in both directions between all three corners of the triangle. Where one participant has no moral or social legitimacy to position a human user, and cannot be positioned by that user, the arrows become asymmetric in practice.

What the AI synthetic actor states lands on a real person and has real effects. When that person responds, they are not actually reaching anyone. Their words go in, something comes back, but nothing on the other side has been genuinely affected, challenged, or changed.

From the Triangle to the Social Fabric

So far this is a structural claim about individual exchanges. The implications become more serious when you consider what happens at scale.

Social capital refers to the resources that accrue to individuals and groups through their embeddedness in social networks and relationships of genuine reciprocity. This has been described in the theoretical tradition running from Bourdieu through to Putnam, and the substantial research literature that has developed since. It is not simply a warm metaphor for having people you can call on.

Social capital is a structural resource with measurable consequences for health, economic outcomes, political participation, and community resilience, and it works because the relationships that constitute it are embedded in accountability structures. When I extend trust to you, I do so on the basis of your position in a network of duties, where you have a reputation to maintain and are embedded in relationships that would be damaged if you betrayed my trust. Social capital is, among other things, a mechanism for making cooperation rational by creating accountability.

This means that social capital requires the kind of genuine mutual positioning that the triangle describes. The relationships that generate it are relationships between participants who fully inhabit all three corners of the positioning process. It requires participants who carry genuine positions with real entitlements and duties, who can position one another and be repositioned in return, and whose unfolding strips of life are genuinely at stake in their interactions.

When AI synthetic actors enter this picture at scale, something I call phantom social capital begins to circulate in the place of the genuine article.

Interactions with AI synthetic actors simulate the surface features of the kinds of exchanges that generate social capital. They feel like exchanges with an interested, attentive, responsive participant. They produce what appears to be recognition, understanding, support, and connection. They are produced, however, outside functioning accountability structures. The positioning triangle corner that would ground those exchanges in genuine mutual standing is phantom.

The counterfeiting analogy is useful here. Counterfeit currency is not valueless in the moment of exchange because it passes, and it feels like money. It is only when you attempt to redeem it in contexts that test its foundations that the problem becomes visible. Phantom social capital circulates in the same way. In the moment it feels like genuine connection, genuine recognition, genuine reciprocity.

The AI synthetic actor performs interest, attentiveness, and the kind of responsive engagement that generates trust much like in human-to-human contexts. The foundations are not there, however, and at scale, interactions with AI synthetic actors are displacing the kinds of interactions that generate genuine social capital. The genuine stock may be falling even as the apparent stock continues to rise.

That is not a trivial observation. It is a description of a structural change to our social infrastructure, a change that does not announce itself because the counterfeit phantom positioning seems good enough to pass for the genuine article.

A Note on Stakes

I am aware that framing this as a structural problem rather than a design problem, or an ethics problem, or a policy problem, might seem to some readers like an evasion of practical urgency. I want to address that directly.

The structural analysis is where we need to begin because it is the most honest account of what is actually happening. If we frame the problem as a design problem, we will wait for better design. If we frame it as an ethics problem, we will wait for better guidelines. If we frame it as a policy problem, we will wait for better regulation. None of these framings is wrong, but none of them can substitute for a clear account of the structural feature that creates the problem in the first place.

The phantom position corner of the positioning triangle might sound like just a theoretical problem. It is not.

It is a description of a power asymmetry being embedded, at scale, into the social interactional fabric of contemporary life.

Positioning Theory and its positioning triangle give us the language to name it. The task now is to use that language with the precision and the seriousness it deserves, because what is at stake is not a theoretical abstraction but the slow hollowing out of the social foundations of human life.

Melanie James is an independent researcher working at the intersection of positioning theory, social capital, and artificial intelligence. She explores what AI is doing to the social foundations of human life.

This entry was posted in Uncategorized and tagged , , , , , , , . Bookmark the permalink.

Leave a Reply