Deepfakes, Signs, and Qualia

If the image is false, why can what we feel be real?

A word appears on the screen:

FALSE.

For a few moments, nothing else happens.

But the Body-Territory is no longer exactly the same Body-Territory that existed before the word appeared.

There may be shifts in attention, recruitment of memories, synaptic activity, autonomic changes, subtle changes in breathing, heartbeat, muscle tension, blood flow, or other physiological dynamics. Some of these may reach conscious experience. Many may not.

Then a photograph of a smiling person appears.

The smile does not encounter the Body-Territory that existed before the word FALSE.

It encounters a Body-Territory that has already lived through that first encounter.

Later, if the word FALSE appears again, it encounters another Body-Territory — one that has now also passed through the smiling face.

We can write:

Body-Territory₀ + FALSE → Body-Territory₁

Body-Territory₁ + smile → Body-Territory₂

Body-Territory₂ + FALSE → Body-Territory₃

The pixels composing FALSE may be rigorously identical in both presentations.

The encounter is not.

The same stimulus can be repeated. The same encounter cannot.

This difference is central to understanding deepfakes, signs, and Qualia — and also to understanding what we call the 5D Materiality of Consciousness — Body-Territory.

A face can remain the same while its meaning changes

In 2023, Anna Eiserbeck and colleagues conducted a particularly interesting experiment.

Participants viewed faces preceded by the words REAL or FAKE. But there was a crucial detail: all the images were photographs of real people. No artificially generated face had been included. The experimental manipulation concerned the belief about the origin of the image, not physical differences between a real face and a deepfake. (nature.com)

Faces classified as fake did not produce exactly the same experience.

In particular, smiles presented as fake were rated as less positive and required slower responses. In EEG, differences normally found between happy and neutral expressions in P1, N170, and EPN appeared when faces were believed to be real, but were reduced or absent when they were presented as deepfakes. LPP dynamics also differed. The authors themselves, however, emphasize limitations: electrophysiological interactions were not significant in every analysis, and the pattern depended on emotional valence. (nature.com)

We should therefore not turn the study into the simple claim that “the word FAKE produces a general priming effect.” In fact, the authors argue that the emotion-dependent pattern works against an explanation based exclusively on generic priming. (nature.com)

The result is more interesting:

The physical smile remained the same. But what the smile was able to mean for the Body-Territory could change.

The encounter begins before the image

In previous texts we introduced the concept of priming.

Here it needs to be understood within a larger structure.

When the word FALSE appears, we are not saying that an abstract entity called “the concept of falsity” simply waits inside some mental compartment.

The previous sign encounters a living organism.

That encounter has materiality.

It may recruit neural networks, memories, autonomic responses, interoceptive states, and proprioceptive states. It can alter the physiological configuration that will encounter the next stimulus.

Therefore:

The next encounter meets the materiality that the previous encounter has already helped configure.

This is the sense in which we can say:

The encounter with the image can begin before the image.

Not because the future has already occurred.

But because the present that will receive that future is continuously being produced.

The Body-Territory was already happening

This formulation brings us closer to the line developed by Antonio Damasio and Gil Carvalho on interoception and homeostatic feelings.

For them, feelings represent bodily states as the organism undergoes homeostatic regulation. Interoception continuously involves neural and humoral signals originating inside the organism — including the viscera and circulating molecules — rather than only information arriving from the external world. (pubmed.ncbi.nlm.nih.gov)

Thus, when the word FALSE appears, it does not encounter merely:

“a brain looking at a screen.”

It encounters hunger or satiety.

Pain or comfort.

Sleep or sleep deprivation.

Breathing.

Heartbeat.

Body position.

Temperature.

Recruitable memories.

Social history.

Expectations.

And everything that, at that moment, materially constitutes that organism.

The world arrives at a Body-Territory that was already happening.

5D Materiality: before Qualia, there is a material Body-Territory

This is where the BrainLatam concept must be made explicit.

In the 5D Materiality of Consciousness — Body-Territory, the first three dimensions do not represent an “abstract mental space.”

They concern the spatial materiality of experience.

A visual perception.

An interoceptive signal.

Proprioceptive information.

A recruited memory.

A word.

An imagined image.

For any of these to participate in experience, they must occur in the present as a material configuration of the Body-Territory.

Depending on the phenomenon and the scale being observed, this may involve:

synaptic activity, neuronal and glial networks, ionic gradients, blood circulation, metabolism, cardiac activity, breathing, muscle tension, and other transformations of the organism.

The remembered tree is not literally inside the brain.

But remembering the tree is materially happening now.

The origin of a representation may be absent. Its instantiation in the Body-Territory must be present.

From the atom to behavior: Triple-Aspect Monism

This is where Alfredo Pereira Jr. provides a particularly important ontological bridge.

In Triple-Aspect Monism, Pereira Jr. proposes three aspects necessary for conscious experience: Matter, Information, and Sentience, derived from a neutral basis called Energy. In his 2025 multiscale formulation, Matter corresponds to energy condensed into particles; Information to differential distributions of energy in space; and Sentience to the capacity to feel, which he relates to temporal energetic patterns in living systems. (neuralpress.org)

This allows BrainLatam to build a multiscale extension without reducing consciousness to neuronal firing:

energy → matter → form/information → biological organization → sentience

and, in living systems:

atoms → molecules → proteins → cells → tissues → organs → neuroglial networks → body → behavior → relation → territory

This does not mean that every one of these levels must be measured in every experiment.

It means that:

The experimental cut may choose one scale; the Body-Territory continues happening across the others.

Pereira Jr. has also proposed an exploratory model of astroglial hydro-ionic waves, involving calcium, other ions, water, proteins, and extracellular matrix in neuroglial modulation. The hypothesis is exploratory and should not be treated as an established mechanism of consciousness, but it is important because it presents water, ions, proteins, and extracellular matrix as more than a passive background to neural activity. (journals.sagepub.com)

The BrainLatam extension can move still further down the scale: DNA expression modifies protein production; proteins constitute channels, receptors, transporters, enzymes, structural elements, and extracellular matrix; these participate in ionic and osmotic gradients, water movement, metabolism, and cellular dynamics.

We are not attributing this specific chain to Pereira Jr.

We are using his ontology to remember that:

Information does not need to float above matter. In living organisms, informational differences can reorganize materiality itself.

Movement: each encounter produces the next Body-Territory

The fourth dimension of 5D Materiality is Movement.

Not merely the abstract passage of time.

Movement is transformation:

BT₀ → BT₁ → BT₂ → BT₃...

One configuration is recruited.

Another diminishes.

A memory returns.

Breathing changes.

Attention shifts.

One representation becomes associated with another.

One process is inhibited.

Another gains strength.

Some changes disappear rapidly through homeostatic regulation.

Others may persist for seconds, hours, or years.

And many may be too small for the instrument used in a particular experiment to detect.

But:

“Negligible for the experiment” does not mean “nonexistent in the Body-Territory.”

An EEG may fail to detect a difference.

A questionnaire may fail to find it.

A statistical comparison may treat it as irrelevant.

That means only that the difference was not shown to be relevant for the chosen variable, in that design, at that resolution, in that sample.

It does not mean that Body-Territory₂ is literally identical to Body-Territory₀.

Then we arrive at Qualia

Within 5D Materiality, Qualia is the fifth dimension.

It does not appear only as a verbal response after the event:

“I liked it.”

“I felt afraid.”

“It seemed strange.”

Qualia is the coloring of experience.

What stands out.

What shines.

What weighs.

What pulses.

What threatens.

What welcomes.

What becomes familiar.

What gains urgency.

Qualia is what makes a material configuration in Movement matter in that particular way for that Body-Territory.

Pereira Jr. also treats qualia as first-person qualitative experiences and distinguishes the scientific investigation of objective patterns from singular access to felt experience. (anpof.org.br)

Here we reach the decisive connection with the deepfake:

The sign may remain physically unchanged while what pulses within it changes.

A smile may continue to smile.

But it may stop shining as closeness and begin to pulse as artificiality, suspicion, or strangeness.

The image can be artificial. Qualia is not therefore “false”

In 2026, Alice Mado Proverbio and Mariia Dosaikina compared real human faces with hyper-realistic AI-generated faces. Participants had difficulty distinguishing the categories, while EEG showed modulations involving N170, N250, P300, PN400, and late positivity. The study involved 30 participants and does not justify speaking of a “brain AI detector,” but it does show that behavioral indistinguishability and electrophysiological processing do not have to be equivalent. (nature.com)

Another study, published in 2024, showed that the same kinds of synthetic faces received lower attractiveness ratings when participants believed they had been produced by AI; that belief also modulated fMRI signals in the fusiform cortex. (doi.org)

We therefore need to separate two realities:

the origin of the sign

and

the reality of the experience produced in the encounter.

A synthetic person may never have existed.

But the change in breathing exists.

The tension exists.

The neural activity exists.

The affect exists.

The Qualia happens.

Falsity may belong to the sign. Feeling belongs to the encounter.

But which humanity did AI learn to make “real”?

The question becomes decolonial when we ask who participates in the statistical definition of a “normal,” “realistic,” or “human” face.

In 2023, Miller and colleagues demonstrated a phenomenon they called AI hyperrealism: AI-generated white faces were classified as human more often than real white faces. The authors related this result, among other factors, to the disproportionate representation of white faces in data used by generative systems, and they made clear that the effect was not equivalent for non-white faces. (psychologicalscience.org)

Brazilian researchers Juliana de Assis and Maria Aparecida Moura analyzed generative-tool images in 2025 from a semiotic and racial perspective, identifying patterns of representation and limited diversity in particular categories. (scielo.br)

Another 2025 study found racial and gender stereotypes and within-group homogenization in images generated by Stable Diffusion; in a preregistered experiment, exposure to inclusive or non-inclusive image sets was also associated with later differences in measures of bias. (nature.com)

Therefore:

When we ask whether an artificial face looks “real,” we must also ask which part of humanity was most heavily used to teach the machine how to represent the real.

NeuroDesafio LATAM research question

As a sponsor of NeuroDesafio LATAM, BrainLatam proposes:

When the same stimulus is presented as human or artificial, how do the Body-Territory’s materiality in Movement and the Qualia attributed to the experience change?

A Latin American experiment could cross:

actual origin: human × AI;

presented belief: human × AI × uncertain.

EEG could track P1, N170, EPN, N250, P300/PN400, and LPP, without turning any of them into a “Qualia meter.”

fNIRS could follow cortical hemodynamic changes during longer evaluations.

But a 5D design should go beyond the brain.

Breathing.

ECG.

electrodermal activity.

potentially movement and muscle tension.

first-person report.

And cultural context.

Because:

EEG and fNIRS are windows onto parts of materiality and Movement. They are not the Body-Territory itself, and they do not directly measure Qualia.

The final question may be even more important:

When the label “FALSE” changes what a smile makes pulse within us, exactly when did the change begin: in the word, the face, the recruited memory, the heart, the breathing, the neural network — or in the multiscale relation among all of them?

Perhaps that is precisely the mistake produced by disciplinary “well-lit places”: searching for consciousness only where our instrument knows how to shine.

5D Materiality proposes another direction:

The instrument chooses a window. The Body-Territory continues happening beyond it.

A screen transduces the cut.

The encounter materializes that cut again in the Body-Territory.

That materiality enters Movement.

And something within it may acquire Qualia.

It may shine.

It may weigh.

It may pulse.

An image may be false. But the Body-Territory that encountered it became materially part of the history of that encounter.

Commented References

Eiserbeck, A.; Maier, M.; Baum, J.; Abdel Rahman, R. (2023). “Deepfake smiles matter less—the psychological and neural impact of presumed AI-generated faces”. Scientific Reports, 13, 16111.
This is the central experimental reference for this text. All faces were real, but half were labeled “FAKE.” Presumed-fake smiles had less behavioral impact and showed differences in analyses involving P1, N170, EPN, and LPP. The authors themselves emphasize limitations and argue against a generic priming-only explanation, which is fundamental to our careful interpretation of the study. (nature.com)

Proverbio, A. M.; Dosaikina, M. (2026). “Neural signatures of hyper-realistic AI-generated faces: dissociating behavioral indistinguishability from implicit neural evaluation”. Scientific Reports, 16, 22944.
The study found modulations in N170, N250, P300, PN400, and late positivity in response to hyper-realistic AI faces, even when participants had difficulty distinguishing them behaviorally. It supports our distinction between what participants can explicitly report and the physiological dynamics accompanying the encounter. (nature.com)

Nishida, S. (2024). “Behavioral and neural evidence for the underestimated attractiveness of faces synthesized using an artificial neural network”. Computers in Human Behavior: Artificial Humans, 2(2), 100104.
Participants rated faces as less attractive when they believed they were synthetic, despite controlled physical appearance, and this belief modulated responsiveness in the fusiform cortex. It reinforces the idea that information about origin participates in the experience generated by the sign. (doi.org)

Carvalho, G. B.; Damasio, A. (2021). “Interoception and the origin of feelings: A new synthesis”. BioEssays, 43(6), e2000261.
This work treats feelings as conscious events related to bodily states under homeostatic regulation and emphasizes the interaction between neural and non-neural components of the organism. It provides the bridge for our formulation that an external stimulus always encounters a Body-Territory that is physiologically already underway. (pubmed.ncbi.nlm.nih.gov)

Damasio, A. (2021). Feeling & Knowing: Making Minds Conscious.
Develops the relationship among life, homeostasis, feeling, knowing, and consciousness. For 5D Materiality, it helps prevent consciousness from being treated as something that begins only when an external stimulus reaches the brain. (penguinrandomhouse.com)

Pereira Jr., A. (2023). “Introdução à Metafísica do Monismo de Triplo Aspecto”. ANPOF.
Presents Matter, dynamic patterns of Information, and Sentience as jointly necessary aspects of conscious experience and connects first-person and third-person perspectives. It provides the philosophical basis for refusing to reduce Qualia to an electrophysiological marker. (anpof.org.br)

Pereira Jr., A. (2025). “The interdisciplinary ontology of consciousness in multiscale neuroscience”. Journal of Multiscale Neuroscience, 4(4), 276–293.
Makes Triple-Aspect Monism explicit as a multiscale ontology: Energy as the basis and Matter, Information, and Sentience as fundamental aspects. It is the ontological bridge used by BrainLatam to think across scales without choosing the neuron as the only legitimate level of description. (neuralpress.org)

Pereira Jr., A. (2017). “Astroglial hydro-ionic waves guided by the extracellular matrix: An exploratory model”. Journal of Integrative Neuroscience, 16(1), 57–72.
Exploratorily proposes interactions among calcium waves, water, ions, proteins, extracellular matrix, and neuronal activity. It is not presented here as a proven mechanism of consciousness, but as an important Brazilian attempt to think neurobiology beyond an exclusively neuronal-electrical description. (journals.sagepub.com)

Miller, E. J. et al. (2023). “AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones”. Psychological Science, 34(12), 1390–1403.
Shows that AI-generated white faces could be judged as human more often than real white faces, a phenomenon that was not equivalently generalized to other groups. It makes the question of “realism” inseparable from the composition of the data used by AI systems. (psychologicalscience.org)

Assis, J.; Moura, M. A. (2025). “Semioses algorítmicas e viés racial: um estudo de imagens criadas pela IA generativa”.
A Brazilian study analyzing generative representations through a semiotic and racial lens. For Decolonial Neuroscience, it helps shift the question from “Can AI produce a realistic face?” toward “Which bodies and repertoires participated in the algorithmic construction of what we call real?” (scielo.br)

AlDahoul, N.; Rahwan, T.; Zaki, Y. (2025). “AI-generated faces influence gender stereotypes and racial homogenization”. Scientific Reports, 15, 14449.
Documents gender and racial stereotypes and within-group homogenization in generative images, and experimentally shows that different representational sets can be associated with later changes in measures of bias. It reinforces the hypothesis that an image does not end when it is seen: it participates in the conditions under which future signs will be encountered. (nature.com)

BrainLatam — 5D Materiality of Consciousness — Body-Territory.
This text consolidates four propositions in the series: “The same stimulus can be repeated. The same encounter cannot”; “The world arrives at a Body-Territory that was already happening”; “Movement is the material history of the Body-Territory happening”; and “Qualia is the coloring of experience — what stands out, shines, weighs, or pulses.” The BrainLatam proposition is transdisciplinary: an experiment may cut into brain, heart, behavior, or language, but none of these windows, in isolation, exhausts the multiscale materiality of the Body-Territory.