Abduction, Prediction, and Inference

When a question, a poll, or an AI lights up the hypothesis before we do

Someone receives a phone call.

Before being asked whom they would vote for, other questions appear:

What problem worries your family the most?
How do you evaluate the economy?
Has public safety improved or worsened?
Whom do you trust to address these problems?

Only afterward comes the electoral question.

We could imagine that the final answer was ready from the beginning, stored inside the person and simply waiting to be extracted.

But research on survey methodology shows something more complex: the order, context, alternatives, and wording of questions can change the responses that are recorded. Earlier questions can make certain considerations more accessible when the next question arrives. The Pew Research Center describes these as order effects and uses randomization precisely to reduce this kind of influence. (Pew Research Center)

Perhaps, then, a survey is not merely finding an answer.

Under certain conditions, it may also participate in the conditions under which that answer is constructed.

A question does not merely illuminate a possible answer. It can illuminate first the region in which the answer will be sought.

This is where priming, abduction, prediction, and Artificial Intelligence begin to meet.

Before abduction, there may already be priming

In Cognitive Psychology, priming describes situations in which a previous stimulus modifies subsequent processing.

A meta-analysis developed at the University of São Paulo examined priming in communication and identified an important role for stimuli associated with goals and motivational systems. The study also emphasized the heterogeneity of experimental findings, reminding us that priming should not be treated as an automatic mechanism of mind control. (University of São Paulo)

In another Brazilian study, Bernardo Fajardo and Guilherme Abib Leão found that previous exposure to concepts related to ethics and fraud modified subsequent judgments about unethical behavior. (SciELO)

The phrase we have been using therefore needs another layer:

The perceived object encounters habits, memories, affects, and representations that may already have been primed by previous encounters.

This matters because the Body-Territory does not encounter each event from zero.

It meets the present carrying a long history — habits, culture, experiences — and an immediate history: what has just been seen, heard, felt, or asked.

The previous encounter can participate in the meaning of the next encounter before we even notice its participation.

Then comes abduction

For Charles Sanders Peirce, when we encounter a surprising fact, we search for a hypothesis capable of making it intelligible.

A glass appears shattered on the floor.

Perhaps it was the cat.

Perhaps the wind.

Perhaps someone entered the house.

Abduction does not demonstrate which answer is true. It opens a possible explanation.

But where do the hypotheses that manage to appear come from?

In BrainLatam's 5D Consciousness model, we propose that an event encounters representational spaces with different degrees of recruitment. Some pathways have been traversed so often that they may reappear quickly. Others are less available.

Priming adds a temporal dimension to this.

A pathway can be strongly recruited because it was consolidated over many years.

Or it may have been illuminated thirty seconds ago.

Thus:

previous stimulus → priming → more available possibilities → new encounter → hypothesis → prediction → action

This approximation between priming and 5D Consciousness is a BrainLatam hypothesis, not an already established neuroscientific equivalence.

But it gives us a simple formulation:

Priming does not determine thought. It changes the terrain on which the next thought may occur.

A survey is also an encounter

This principle changes how we can think about public opinion research.

Pew shows that apparently small changes in question wording can produce substantial differences in responses and that earlier questions can generate assimilation or contrast effects on later answers. (Pew Research Center)

In our reading, the sequence can be:

previous question → priming → new question → search for meaning → hypothesis → response

Not every order effect is technically “abduction.” Survey methodology uses concepts such as question-order effects, framing, priming, assimilation, and contrast.

The BrainLatam contribution is to ask what happens after priming.

When encountering a new question, the Body-Territory may need to construct an interpretation:

“If they just asked me about X and now they are asking me about Y, perhaps X is relevant to understanding Y.”

This has an abductive character.

The first question may not provide the answer. It may provide the region in which the next answer will be sought.

When the poll result returns to the voter

There is also a second circuit.

A poll interviews people.

It transforms responses into numbers.

The numbers are published.

And they return to the population as new signs.

At that moment, the poll ceases to function only as a measurement instrument and also becomes information circulating within the electoral environment.

The evidence does not show one single inevitable effect. A large Dutch experiment involving 23,421 participants did not find a simple change in vote intention merely from presenting poll numbers, but it did find an effect when emphasis was placed on a particular option gaining support. (Oxford Academic)

Danish experiments, on the other hand, found evidence of a bandwagon effect: information indicating increasing support was followed by greater intention to support the same option. (Wiley)

In Latin America, Rodrigo Uribe and Enrique Manzur, from the Universidad de Chile, experimentally examined the influence of opinion polls on preferences and found effects associated with the strength of the results presented, under certain individual conditions. The authors themselves discuss polls as possible instruments of influence, while noting that the literature does not show universally uniform effects. (SciELO Chile)

The scientifically appropriate conclusion is therefore not:

“Election polls control the vote.”

It is:

The publication of polls can, under certain conditions, become part of the informational environment that participates in the formation or updating of preferences.

That is already significant.

Is an election poll political advertising?

Here we need to separate a scientific question from a legal one.

From a semiotic perspective, the publication of a poll can function as a persuasive sign:

“This candidate is growing.”

“This option is leading.”

“This alternative appears viable.”

But that does not automatically make every electoral poll legally equivalent to political advertising.

In Brazil, election polling in 2026 is subject to its own regulatory framework. Law No. 9,504/1997 and the rules consolidated by Brazil's Superior Electoral Court, including the 2026 amendments, require prior registration in the PesqEle system and information such as the contracting party, source of funding, methodology, sampling plan, margin of error, confidence level, and the complete questionnaire. (Brazilian Federal Legislation)

This is especially significant for our argument:

If wording and order can change responses, the complete questionnaire is also part of the scientific data needed to interpret the result.

Knowing only the final percentage is not enough.

The Brazilian state already has oversight mechanisms

The rules applicable to Brazil's 2026 elections require polls intended for public release to be registered at least five days before publication. Registrations can be consulted, and authorized parties may request access to methodological control elements and the questionnaire. Electoral courts do not conduct prior approval of poll results, but when legally challenged and the applicable requirements are met, there may be judicial review, suspension of publication, or orders for clarification. (TSE)

Publishing an unregistered poll can lead to fines, while knowingly publishing a fraudulent poll can constitute an electoral offense under Brazilian law. (TSE)

There is also an important distinction between an informal poll or straw poll and a methodologically structured electoral survey. Brazilian electoral regulation restricts informal electoral polling under specified conditions, while registered surveys can be published when they comply with the applicable requirements.

This shows that Brazilian law already recognizes that information presented as electoral measurement requires rules of transparency, traceability, and oversight.

A separate question is what methodological controls should apply to wording, question order, priming, or the way results are communicated. That is a regulatory choice involving electoral integrity, freedom of information, scientific independence, and auditability.

Survey methodology, Semiotics, and Neuroscience can provide evidence about these effects.

They do not, by themselves, determine what the legal solution should be.

When AI enters before the hypothesis

Now we return to generative AI.

A person encounters a problem and immediately asks:

“What is the most likely explanation?”

The AI offers a hypothesis.

Even if the person later rejects it, that possibility has already entered the representational space.

Before persuading, an AI can prime.

Experiments on creativity reveal this ambivalence. Doshi and Hauser found that AI suggestions improved some evaluations of individual creativity, particularly among participants who initially scored lower in creativity, but they also made the final productions more similar to one another. (PubMed)

This raises a decisive question:

Can an intelligence expand my individual possibilities while simultaneously moving my possibilities closer to those it offered everyone else?

EEG research has already shown that self-generated and externally induced expectations can produce differences in markers such as N2 and P3. (Frontiers)

Research published in 2026 has also reported electrophysiological differences between self-generated and externally instructed action sequences, including readiness potentials and beta suppression. (eNeuro)

This does not mean that we know a “neural signature of abduction.”

It means we have tools for asking whether the origin of a hypothesis matters for what happens next.

NeuroDesafio LATAM research question

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

Does a hypothesis previously suggested by AI modify the generation, selection, or updating of the hypotheses subsequently produced by the participant?

EEG could track expectation, conflict, and updating with fine temporal resolution. fNIRS could monitor prefrontal recruitment during longer periods of deliberation.

A second paradigm could investigate something equally important: presenting questions in different orders before an ambiguous problem and examining whether prior activation changes the hypotheses participants subsequently generate.

Not to discover “how to control an opinion.”

But to understand:

How much of the hypothesis we experience as our own had already begun to gain availability before we realized we were choosing?

Autonomy may not require a mind untouched by influence.

It may require something more realistic:

preserving enough time and space so that the first hypothesis illuminated does not automatically become the only visible reality.

Because a screen has light.

A question does too.

A poll does too.

An AI does too.

And what we illuminate first can participate in what we are able to see next.

Commented References

1. Pew Research Center. “Writing Survey Questions”.
   Documents experimentally that wording, question order, and response-option order can change survey responses. It is fundamental to our proposition that a survey does not necessarily encounter an opinion under neutral conditions; the questionnaire itself can provide the context within which that opinion is formulated. (Pew Research Center)

2. Hughes, A.; Jones, B. (2019). “‘Good jobs’ vs. ‘jobs’: Survey experiments can measure the effects of question wording – and more”. Pew Research Center.
   Shows how randomizing wording and order can causally identify effects generated by the survey instrument itself. It supports the phrase: “the order of questions is also an order of possible recruitments.” (Pew Research Center)

3. Fajardo, B. A. G.; Leão, G. A. (2014). “O efeito priming na avaliação de ações antiéticas: um estudo experimental”. Revista de Administração Contemporânea, 18(1), 59–77.
   This Brazilian study experimentally demonstrates how prior activation can participate in subsequent judgments. It provides a Latin American bridge among priming, judgment, and our hypothesis of representational recruitment. (SciELO)

4. Senise, D. S. V. (2015). Efeito priming aplicado em comunicação: uma meta-análise. University of São Paulo.
   Reviews national and international studies of priming in communication, emphasizing goals and motivational systems while also documenting variability in effect sizes and experimental results. This helps prevent “priming” from being transformed into a synonym for inevitable manipulation. (USP)

5. Uribe, R.; Manzur, E. (2007). “Los Estudios de Opinión y su Influencia en las Preferencias de las Personas”. Psykhe, 16(2), 97–105. Universidad de Chile.
   A Latin American reference directly related to electoral polling. The authors examine bandwagon and underdog effects and present experimental evidence that characteristics of polling information can be associated with changes in preference. The study illustrates how a published poll result can return to the public as a new social stimulus. (SciELO Chile)

6. Van der Meer, T. W. G.; Hakhverdian, A.; Aaldering, L. (2016). “Off the Fence, Onto the Bandwagon? A Large-Scale Survey Experiment on Effect of Real-Life Poll Outcomes on Subsequent Vote Intentions”. International Journal of Public Opinion Research, 28(1), 46–72.
   In a study involving 23,421 participants, poll results alone did not produce a simple uniform change in vote intention, although emphasizing that an option was gaining support did have an effect. This is important because it shows that poll influence is contingent rather than automatic. (Oxford Academic)

7. Dahlgaard, J. O.; Hansen, J. H.; Hansen, K. M.; Larsen, M. V. (2017). “How Election Polls Shape Voting Behaviour”. Scandinavian Political Studies.
   A Danish national experiment found evidence of a bandwagon effect when participants received information showing increasing or decreasing support. Together with the previous study, it shows that under some conditions, the trajectory presented by a poll can participate in preference updating. (Wiley)

8. Brazil. Law No. 9,504/1997, Article 33.
   Establishes registration requirements for electoral polls intended for public dissemination, including information about contracting parties, funding, methodology, sampling, and questionnaires. It provides the legal basis for treating electoral polling as an activity subject to specific transparency requirements. (Planalto)

9. Brazilian Superior Electoral Court — Resolution No. 23.600/2019, consolidated with subsequent amendments applicable to the 2026 elections.
   Details the PesqEle registration system, disclosure, inspection, access to questionnaires, legal challenges, and sanctions. Electoral courts do not pre-approve poll results, but applicable irregularities may be subject to judicial review. (TSE)

10. Doshi, A. R.; Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content”. Science Advances, 10(28), eadn5290.
    Experimentally demonstrates a tension highly relevant to abduction: generative suggestions can improve individual productions while also increasing their collective similarity. It supports the question: when AI offers the first possibility, does it expand the space of subsequent hypotheses — or also anchor it? (Science Advances)

11. Kemper, M. et al. (2012). “What I Say is What I Get: Stronger Effects of Self-Generated vs. Cue-Induced Expectations in Event-Related Potentials”. Frontiers in Psychology, 3, 562.
    Self-generated expectations produced different behavioral and electrophysiological effects from cue-induced expectations, including N2 and P3 differences when expectations were violated. It provides a precedent for investigating self-generated hypotheses versus hypotheses previously suggested by AI. (Frontiers)

12. Seghezzi, S.; Haggard, P. (2026). “Neural Mechanisms of Self-Generated Action Sequences”. eNeuro, 13(5).
    EEG distinguished self-generated from externally guided actions through readiness potentials, beta dynamics, and multivariate analyses. It does not demonstrate a neural signature of abduction, but it shows that self-generation and external guidance can produce distinguishable neural trajectories. (eNeuro)

13. BrainLatam — 5D Consciousness, Priming, and Body-Territory.
    This text consolidates the hypothesis: “The Body-Territory meets the present carrying its long history and the priming of its immediate history.” It adds a consequence: “Once something lights up a hypothesis, we need to ask how long that light continues to participate in what we are able to see.” These are BrainLatam hypotheses intended for experimental investigation, not established neuroscientific mechanisms.