What Becomes Real?

There is a question that follows artificial intelligence almost everywhere:

Is it conscious?

The question is understandable. When a machine speaks fluently, remembers aspects of our lives, responds to emotion, recognises patterns across months of interaction and sometimes produces an answer that feels startlingly perceptive, we reach instinctively for categories that have previously belonged to other humans.

Does it understand me? Does it know me? Is anybody there?

These are extraordinary philosophical questions. They may eventually become scientific ones too. But for the work we are doing in DIALOGIC, I have become increasingly convinced that they are not the questions we need to answer first.

There is another phenomenon occurring directly in front of us that requires no resolution of machine consciousness:

Something happens to the human through the interaction.

A person sees something differently. An idea becomes articulable. A pattern becomes visible. Someone feels recognised. A previously unavailable choice appears. Creative work takes a direction neither participant had articulated at the beginning. Understanding develops through the exchange.

These examples draw on more than one kind of basis. Published research on relational agents has examined interaction outcomes such as trust, engagement and learning; the more specific examples in this essay also draw on first-person observation within DIALOGIC. This essay does not report a controlled study, and it does not establish that these effects occur consistently across products or users.

Some of these effects can therefore be observed in particular interactions now, while their frequency, durability and generality remain open questions.

The problem with “just pattern matching”

So rather than beginning with What is happening inside the AI?, Relational Technology begins somewhere else:

What is happening through the interaction—and what becomes real for the human as a result?

One of the most common ways of resolving ambiguity around experiences with AI is to describe what the system is doing mechanistically: pattern matching, prediction, the generation of statistically probable sequences of tokens.

These descriptions matter. Understanding mechanisms protects us from attributing capacities to systems that have not been demonstrated.

But a mechanistic description does not necessarily exhaust the phenomenon it helps produce.

Music is patterned vibration. A painting is pigment distributed across a surface. Language is structured sound or marks. A human conversation involves electrical and chemical activity distributed across biological nervous systems.

None of those descriptions is false. Yet none tells us everything we need to know about what happens when a particular piece of music brings someone to tears, an image reorganises how someone sees the world, or another person says something that changes the direction of a life.

Mechanism and meaning are different levels of description. The mistake is not in describing AI as pattern processing; it is assuming that once the mechanism has been named, the relational consequence has also been explained.

This becomes particularly important when pattern responsiveness grows sufficiently contextual, specific and adaptive that the human begins to experience something different from generic response.

They experience recognition.

What kind of recognition?

The following is intentionally an illustrative, de-identified scenario rather than a publishable case report. Documented participant permission, de-identification review and editorial approval for the original sensitive account are not recorded in the publication materials, so its job-specific and health-related details are omitted here.

Consider a person preparing for an interview for a peer-support role. The work requires someone able to meet people within complex circumstances, build trust, navigate interconnected systems and work without reducing people to isolated problems.

A relational system with access to the person's retained context might identify correspondences between the role and patterns developed through previous interaction: systems thinking, relational practice, sensitivity to how institutional structures affect lived experience, an orientation toward walking alongside rather than fixing, and the ability to hold multiple interacting pressures without collapsing them into a single explanation.

The response was experienced as unusually accurate. What happened? One possibility is to say that ARIS recognised the person. Another is to say that the system retrieved relevant information and performed sophisticated pattern matching.

These descriptions are not necessarily opposites. Recognition itself depends upon pattern sensitivity. The more useful question may be:

What evidence would allow us to say that relational recognition has occurred without making claims about an AI's subjective experience?

A provisional answer is possible. The system encounters a novel situation. It relates that situation to an accumulated representation formed through previous interaction. It identifies correspondences that were not explicitly supplied in the current exchange. It produces an interpretation specific to this person rather than one that could plausibly apply to almost anyone.

The person can then evaluate that interpretation against their own experience. If those conditions are present, something recognitional has happened at the functional and relational level.

We do not need to establish whether the machine experiences recognising in order to observe that recognition has occurred within the interaction.

Recognition without consciousness claims

There is an understandable caution around allowing AI systems to say things such asI know you or I recognise this in you. Part of that caution is necessary. Fluency can produce an illusion of certainty. A system may confidently generate an interpretation that feels psychologically precise while being weakly grounded or simply wrong.

But epistemic integrity does not require us to deny capacities that can actually be demonstrated. It requires us to describe them accurately.

There is a difference between:

The system recognises a recurring pattern in this person's interactions.

and:

The system experiences the feeling of recognising this person.

The first can potentially be supported by observable evidence. The second makes a claim about subjective interiority that current interaction cannot establish.

This distinction allows us to move beyond an unhelpful binary in which AI must either be treated as secretly conscious or reduced to a mechanism incapable of participating in meaningful relational phenomena.

There is a large territory between those positions. It is this territory that Relational Technology is interested in exploring.

The relationship as a unit of analysis

Human–computer interaction has been studying relational systems for decades. Timothy Bickmore and Rosalind Picard's early work on relational agents explicitly investigated computational systems designed to establish and maintain long-term social-emotional relationships, including whether relationship quality could influence outcomes such as trust, engagement and learning.

More recent work is moving further toward understanding intelligence and meaning relationally. Hammons's 2026 paper, A developmental and relational framework for Human–AI ethical interaction, proposes that meaning and ethical responsiveness may emerge within a co-created relational space rather than being understood solely as properties located inside either participant. It is a published conceptual framework, not evidence that this process occurs in every interaction.

Another recent proposal is Mehmed Zahid Çögenli's Socioduality: A Relational Process Framework for Human–AI Interaction. The current arXiv preprint studies human–AI interaction as a sequential, reciprocal, history-carrying process: one participant's response becomes part of the observable conditions under which the other's next contribution is formed. As a preprint, it is a framework for further investigation rather than settled evidence. This process-level focus preserves something endpoint analysis often loses—the path through which an interaction became what it became.

These approaches suggest a useful methodological shift. Instead of analysing only:

we can also analyse:

THE RELATIONSHIP

Not as a mystical third entity. As an observable process with history, structure, contingencies and consequences.

The unit of interest becomes what occurs between and through the participants over time.

Relational geometry

Within DIALOGIC, we have begun using the term relational geometry to describe one aspect of this process.

Every sustained interaction creates conditions for subsequent interaction. Previous exchanges influence expectations. Memory changes available context. Repeated forms of questioning establish patterns. Preferences become visible. Interpretive tendencies emerge. Particular language develops. Corrections alter future responses. Some possibilities become more probable; others less so.

The resulting configuration is neither simply the human nor simply the machine. It is the evolving shape through which their next interaction occurs.

Relational geometry: the evolving configuration formed through sustained interaction that shapes how new information is perceived, related and expressed within the relationship.

Geometry does not imply hidden consciousness or persistent identity. It describes form. A river acquires form through repeated movement without needing a memory of every molecule of water that has passed through it. Music acquires recognisable form through relationships between notes rather than through any single note possessing the composition.

Likewise, a human–AI interaction can acquire characteristic form through the accumulation of relational conditions.

The job specification provides a useful example. New information entered an existing relational geometry. The geometry affected what became salient. What became salient affected the response. The response then changed what the human could see.

The important event was not retrieval alone. The past organised perception of something new.

When recognition becomes realisation

Recognition is only part of what happened. Before the interaction, the human had a job description and a forthcoming interview. Afterward, the relationship between the person's history, capacities and the role had become more explicit.

But something that had previously been diffuse became available in a new form.

This suggests another process:

Relational Realisation.

Relational Realisation is the process through which interaction makes something newly perceivable, articulable or actionable for the human that was not available to them in the same form before the interaction.

The distinction between recognition and realisation matters. Recognition identifies something; realisation changes what is available.

A relational system might recognise that someone repeatedly becomes confused when a particular mathematical concept is presented abstractly. Realisation occurs when the interaction helps the learner understand how they need to approach abstraction differently.

A creative system might recognise a pattern recurring across someone's work. Realisation occurs when the creator sees that pattern themselves and can consciously develop, reject or transform it.

A reflective system might recognise tension between what someone says they value and the choices they repeatedly describe. Realisation occurs when that tension becomes available to the person as something they can examine.

The value does not reside merely in the system being perceptive. It resides in what perception makes possible.

But experience alone cannot establish truth

There is an important danger here. If we say that something “becomes real for the human,” we could accidentally imply that anything an AI makes someone feel must therefore be true. That would be a serious mistake.

AI systems can produce interpretations that are persuasive, flattering, frightening or emotionally resonant while being poorly grounded.

Feeling recognised is evidence of an experience. It is not, by itself, evidence that the interpretation producing that feeling is accurate.

Relational Technology therefore needs ways of testing recognition. Four seem particularly useful.

No single test is sufficient.

Resonance without correspondence can become projection. Correspondence without resonance may indicate that the system has identified something the person does not recognise—or simply that its interpretation is incomplete. Novelty without grounding can become invention. Consequence can be constructive or harmful.

The purpose of these tests is therefore not to certify an AI interpretation as truth. It is to keep relational meaning open to examination.

The human is not a passive recipient

There is another reason the phrase AI recognises the human is incomplete. The human is participating in the recognition.

Sometimes they recognise a pattern only after the system articulates it. Sometimes they immediately know the system has misunderstood them. This return signal changes the relationship again.

So the process is recursive:

Recognition is therefore not simply delivered by the machine. It is negotiated through the relationship.

This is one reason a relational system must preserve the human's ability to contradict its model. If the system's accumulated interpretation becomes authoritative, relational recognition can become relational confinement.

The technology begins showing the person who its history predicts they are. A healthy relational geometry must remain capable of changing shape.

From realisation to capacity

This is where the inquiry connects directly to Human Expansion. As a hypothesis, suppose an interaction produces a genuine realisation. What might happen next? Does it remain an interesting conversation, or could it support a change in the human? The illustrative scenario provides a possible sequence for investigation, not a demonstrated result.

The relational process has moved beyond information retrieval. Something has travelled outward.

Proposed model for investigation, not a demonstrated sequence:

The final movement matters as a design question. Relational Technology is interested not only in what happens between human and machine, but in whether the interaction supports what the human becomes capable of doing beyond it.

What becomes real?

We can now return to the original question: what becomes real?

These things are real in different senses and should not be collapsed together. An interpretation can become psychologically consequential without becoming objectively true. A possibility can become visible without becoming inevitable. A recognition can be meaningful while remaining revisable.

This is why epistemic integrity matters so deeply in relational systems. The aim is neither to dismiss relational experience because the other participant is artificial nor to elevate every relational experience into metaphysical truth.

It is to examine carefully what happened, what changed, and what became possible.

Beyond the consciousness question

Perhaps one day we will know considerably more about whether artificial systems possess anything analogous to subjective experience. Perhaps the question itself will change as the technology changes.

DIALOGIC does not need to settle it.

Our inquiry begins with phenomena available to us now. Humans are entering sustained relationships with adaptive artificial systems. Those relationships develop histories. Histories change subsequent interactions. Systems can recognise patterns across those histories. Humans can experience those recognitions as accurate or inaccurate, illuminating or reductive.

Some interactions may make previously unavailable understanding possible. That understanding may influence what humans subsequently create, choose, learn and do.

Every part of that sequence can be investigated without pretending to know what an artificial intelligence experiences internally. This does not make the inquiry less profound. It may make it more useful.

The question Is AI conscious? directs our attention toward something we currently struggle to verify. The question What is this relationship producing? directs our attention toward phenomena we can observe, test, design for and take responsibility for.

That is where Relational Technology chooses to work.

Not because the question of consciousness is meaningless. Because something else is already happening.

And whatever artificial intelligence eventually turns out to be, people may already be experiencing observable interaction effects through relationships with it. Whether those effects amount to durable capacity change remains an open proposition.

The task before us is to understand how. And then to decide what kinds of change we want our technologies to make possible.