AI, Habits, and Autonomy

Does your algorithm know you - or is it teaching you to keep being who you already were?

You unlock your phone.

You had not planned to open a social network.

And yet, a few seconds later, your finger has already found the app. You scroll. You stop at one video. Then another. You like one. Ignore another. Stay a few seconds longer on a third.

None of these actions seems extraordinary.

But each one leaves a trace.

And the system learns.

The next time, the platform may not simply show you what you like. It may show you what you have already learned to stop for.

This raises an important question:

Does your algorithm know your habits, or does it participate in constructing the very habits it later uses to know you?

It is within this circuit that autonomy, Artificial Intelligence, and Semiotics begin to meet.

Habit does not mean the absence of autonomy

Maria Eunice Quilici Gonzalez and colleagues offer a particularly important entry point into this problem. In examining habits and rationality in the age of Big Data, the authors avoid two extremes: we are neither completely free from our histories and dispositions, nor are we simply puppets of digital technologies.

The hypothesis is more interesting: people may act rationally in many situations while, at the same time, having their opinions influenced by insufficient or distorted information, previous habits, and previously acquired emotional dispositions.

This means that autonomy does not require living without habits.

That would be impossible.

Habits save time and allow us to repeat actions that do not need to be reconstructed from scratch every moment.

The problem appears when a particular path becomes so available that other possibilities almost stop competing for attention.

At BrainLatam, we have been using a simple image:

A familiar branch has been found.

Faced with a new event, the Body-Territory may quickly find an already-built way of perceiving, feeling, interpreting, and acting.

The novelty arrived.

But it encountered an old path.

The network can learn which branch you usually choose

This is where something new appears compared with habits formed before digital platforms.

The environment is also learning from us.

The algorithm observes:

where we stop;
what we repeat;
what we ignore;
who we interact with;
how long we remain;
what kind of expression triggers a response.

Then it reorganizes the environment we will see again.

A circuit emerges:

habit → action → data → algorithm → recommendation → new action → reinforcement or alteration of the habit.

Anderson Vinícius Romanini and Marcelo Hamdan Alvim analyze precisely this relationship between algorithmic mediation, cognition, semiotics, and Active Inference. In their work, algorithms can participate in users' inferential processes, influencing patterns of response and interaction.

This does not mean that an algorithm inevitably determines what someone will think.

Costa Rican researcher Ignacio Siles, in studying users of Netflix, Spotify, and TikTok in Costa Rica, reaches a more balanced conclusion: people incorporate algorithms into everyday life, follow recommendations, resist them, modify their practices, and sometimes do all of these things at once. Agency appears in this intermediate space between user and system.

Perhaps, then, the algorithm is neither simply outside us nor merely our controller.

It enters the relationship.

And the relationship changes both sides.

Not every repeated use is “addiction”

This distinction matters.

In 2025, Ian Anderson and Wendy Wood, a leading researcher on habits, showed that users may overestimate the extent to which their social media use corresponds to clinical dependence. In a sample of active Instagram users, habitual behavior was far more common than the indicators used to identify addiction risk.

The authors emphasize that habits can be automatically activated by contextual cues — opening, scrolling, posting, reacting — without this being equivalent to clinical addiction.

This distinction matters directly for autonomy.

If every repetition is described as irresistible, we may end up reducing our own perception of the capacity to change.

Habit, by contrast, allows a different question:

What cue is activating this response — and can we reorganize our relationship with it?

This question is very close to what we want to investigate through 5D Consciousness.

Stone-Connectome: when the familiar arrives first

In BrainLatam's operational cartography, we use Rock–Paper–Scissors as a metaphor for different functional modes.

Rock represents rapid responses, automatisms, habits, and the reproduction of familiar pathways.

Scissors represents greater analysis, classification, and deliberate planning.

Paper represents openness, flow, and metacognition.

We do not treat these names as three independent anatomical networks, nor as neuroscientific diagnoses. They are an operational metaphor for formulating experimental questions.

In the 5D hypothesis, when certain representational spaces are recruited very frequently, they may become increasingly likely to participate again in experience and may temporarily make it harder for less available alternatives to be recruited.

We can summarize this as:

The more familiar the path, the less effort it requires to be found again.

And formulate an even stronger hypothesis:

The more rigid the already recruited path, the fewer possibilities the new sign may encounter before action occurs.

This is exactly where EEG and fNIRS can help us transform a metaphor into a scientific question.

A notification can already modify the next moment

An electrophysiological study published online in 2023 and in the 2024 issue of Biological Psychology investigated the effect of smartphone notifications during an attention task.

Using EEG, the researchers observed that trials preceded by notifications showed lower theta power and higher alpha and beta power, alongside changes in indicators related to attentional shifting and cognitive control.

A brief mindfulness intervention also modulated some of these effects.

The study does not show that “notifications destroy attention.”

It shows something more useful:

A small digital sign can measurably alter the neural state in which the next stimulus will be processed.

This connects directly with our idea of encounter.

The previous stimulus participates in the conditions under which the next one is processed.

fNIRS is beginning to follow behavior outside highly artificial tasks

In 2025, a study published in Scientific Reports used wearable fNIRS to investigate the immediate effects of social media consumption in 20 university students.

After a brief exposure, participants showed lower accuracy in executive-function tasks, accompanied by changes in prefrontal activity, including regions related to working memory, monitoring, and inhibitory control.

Because of the small sample size, the study should be considered preliminary evidence, not a definitive conclusion about social media users in general.

But methodologically, it is highly relevant.

It shows that NIRS can help investigate digital habits under conditions that are progressively closer to real life.

The algorithm can also help make habits more flexible

It would be a mistake to conclude that algorithmic personalization can only narrow choices.

In a preregistered experiment with 6,488 participants, recommendation systems that learned the preferences people said they wanted to have, rather than only their current preferences, produced slightly fewer clicks but led participants to evaluate their time spent more positively and perceive the system as more aligned with their long-term interests.

This opens an important possibility for our future Body-Territory AI.

Instead of asking only:

“What will make this person click again?”

it could ask:

“Does what I am offering reinforce only the most recruited pathway, or does it help the individual access other possibilities that they themselves consider important?”

In that case, AI would stop being merely a machine for recognizing habits.

It could become a tool for noticing them and making them more flexible.

Decolonial Neuroscience: habits also have territory

We should not assume that a digital habit observed in North American, Chinese, or European students will have exactly the same dynamics in Latin America.

Ignacio Siles and colleagues warn against simply importing theories developed in the Global North and applying them directly to Latin American platforms — a process they critically describe as “tropicalization.”

The Body-Territory using TikTok in San José, São Paulo, Bogotá, an Amazonian community, or Goioerê does not arrive at the screen with the same history.

Language changes.

Infrastructure changes.

Work changes.

Belonging changes.

Inequality changes.

Access to information changes.

Community relationships change.

A Decolonial Neuroscience of digital habits therefore needs to relate EEG or fNIRS not only to the number of clicks, but also to the territory in which that habit acquired meaning.

NeuroDesafio LATAM research question

As a sponsor of NeuroDesafio LATAM / NeuroChallenge LATAM, BrainLatam proposes:

Do algorithms that repeatedly present content congruent with previous habits progressively alter theta, alpha, or measures of attentional and prefrontal control recorded by EEG and fNIRS?

A longitudinal Latin American study could compare people exposed to three types of feeds:

content congruent with previous habits;
deliberately diversified content;
content aligned with the preferences the individual themselves wishes to develop.

EEG could track theta, alpha, ERPs, and attentional-control markers. fNIRS could track prefrontal changes during inhibition and decision-making tasks. First-person reports could help determine whether participants experience a greater or lesser availability of alternatives.

The goal would not be to prove that “the algorithm controls the brain.”

The more interesting question would be:

When a digital environment repeatedly encounters the same branch, does that path become progressively easier to recruit — while the others become harder to notice?

Autonomy may not mean living without habits.

It may mean being able to notice when a habit has started choosing before we do.

AI does not have to learn only which path we usually follow. It can be designed to help us notice that other paths still exist.

Commented References — a second reading of Blog 03

1. Gonzalez, M. E. Q.; Broens, M. C.; Quilici-Gonzalez, J. A.; Kobayashi, G. (2023). “Hábitos e racionalidade: um estudo filosófico-interdisciplinar sobre autonomia na era dos Big Data”. Trans/Form/Ação, 46, 367–386.
   This is the central reference for the blog because it avoids the false opposition between total freedom and complete manipulation: habits and prior information can influence decisions without necessarily eliminating the possibility of relatively autonomous action.

2. Alvim, M. H.; Romanini, A. V. (2025). “Mediações algorítmicas e cognição: conexões entre a semiótica e a inferência ativa”. Esferas, 32.
   This work shows that algorithms do not merely observe behavior; they can participate in the inferential circuit that generates future responses. It supports our chain: habit → data → algorithm → new stimulus → new habit.

3. Siles, I. (2023). Living with Algorithms: Agency and User Culture in Costa Rica. MIT Press.
   This research is especially important for a Latin American approach because it shows users in Costa Rica following, negotiating, and resisting platform recommendations. It helps preserve our concept of agency: people are not passive recipients of algorithms.

4. Siles, I.; Valiati, V.; Valerio-Alfaro, L.; Ferreira, A. (2025). “Tropicalizing platformization? Tensions in research on algorithms and platforms in Latin America”. International Journal of Cultural Studies.
   The article warns against automatically importing concepts from the Global North to explain Latin American experiences. For Decolonial Neuroscience, this means that digital habits need to be investigated within the social and territorial conditions in which they emerge.

5. Anderson, I. A.; Wood, W. (2025). “Overestimates of social media addiction are common but costly”. Scientific Reports, 15, 39388.
   The study distinguishes habit from clinical dependence and shows that automatically labeling frequent use as “addiction” may reduce perceived control. It reinforces one of this blog's central arguments: automatism does not necessarily mean the complete absence of autonomy.

6. Upshaw, J. D.; Shields, G. S.; Judah, M. R.; Zabelina, D. L. (2024). “Electrophysiological effects of smartphone notifications on cognitive control following a brief mindfulness induction”. Biological Psychology, 185, 108725.
   Using EEG, this study found changes in theta, alpha, and beta following smartphone notifications. It provides measurable evidence that a preceding digital signal can alter the neural conditions in which the next stimulus is processed.

7. Aitken, A. et al. (2025). “Naturalistic fNIRS assessment reveals decline in executive function and altered prefrontal activation following social media use in college students”. Scientific Reports, 15, 36960.
   This small wearable-fNIRS study found changes in performance and prefrontal activation after social media exposure. It is important because it shows how NIRS can bring the study of digital habits into increasingly naturalistic conditions, although the small sample requires caution in generalization.

8. Kang, J. et al. (2023). “Tailoring recommendation algorithms to ideal preferences makes users better off”. Scientific Reports.
   This experiment showed that algorithms do not need to optimize only current preferences and clicks: recommendations oriented toward participants' self-declared ideal preferences produced more positive evaluations of time spent. It opens the possibility of imagining AI as a tool for autonomy rather than merely for repetition.

9. BrainLatam — “Neuroscience Perception Avatars” and “Qual é sua Pergunta? — Avatares Neurocientíficos Referenciais” (2026).
   These texts consolidate Rock–Paper–Scissors as an operational metaphor rather than literal anatomy: Rock represents rapid reproduction of the familiar; Scissors, deliberate analysis; Paper, openness and metacognition. Blog 03 turns this cartography into a testable question: does algorithmic repetition make some pathways progressively more available than others?

10. BrainLatam (2026). “O que não entra na atenção também pode limitar o futuro que conseguimos imaginar”.
    This text develops the hypothesis that successively confirmatory stimuli may encounter already available responses and temporarily reduce the accessibility of alternatives. It provides conceptual continuity for the central phrase of this blog: “a familiar branch has been found” — but autonomy also depends on noticing that other branches remain possible.