Generative AI: new cognitive weapon, new alphabet or new common?

Redistributed power

There was a time when technical power was measured in arm strength. Then it shifted to mastery of writing, printing and the machine. Today, it lies in a less visible but more decisive space: the production of language itself. Generative AI doesn't simply add a tool to our digital environment; it intervenes at the heart of symbolic activity. It writes, reformulates, synthesizes and argues. It participates in the making of meaning.

This ability is not neutral. Any technology that amplifies the power to act redistributes social balances. Like the sword in the past, or the printing press later, generative AI reduces certain costs - in this case, cognitive costs - while creating new asymmetries. Those who know how to dialogue with these systems, structure their queries and interpret their answers, see their intellectual productivity multiplied. Those who use them without understanding their mechanisms risk becoming dependent on them. Technical access is not enough: power lies in control.

This raises a decisive question: are we witnessing the emergence of a new cognitive divide? Is generative AI simply an instrument for individual optimization, or is it structurally transforming relationships to knowledge, writing and intellectual authority? In other words, does it become a cognitive weapon, concentrated in a few hands, or can it become a shared alphabet, integrated into learning and placed at the service of an educational common ground?

These questions prompt us to reflect on power, literacy and responsibility. For the emancipatory or inegalitarian effect of generative AI depends less on its existence than on the pedagogical, cultural and institutional conditions of its appropriation. Schools, universities, educational engineers and public decision-makers are not mere spectators: they are called upon to shape the meaning of this technology.

Interrogating generative AI is therefore not just a matter of questioning its uses. It means analyzing the type of cognitive power it institutes, and deciding collectively whether we want it to remain concentrated or become sharable.

AI as a cognitive weapon: amplification and asymmetry

If all technology is an amplification of power, generative AI is a special kind of amplification: it acts directly on higher cognitive operations. It doesn't increase muscular strength or raw computing speed; it intervenes in the formulation, analysis and structuring of language and ideas.

In this respect, it deserves to be called a cognitive weapon, not in the military sense, but in the strategic sense: it modifies the balance of power in symbolic space.

Reducing cognitive costs

First and foremost, AIg acts as a powerful mechanism for reducing cognitive costs. Where information retrieval required exploration, selection and prioritization, it offers immediate synthesis. Where drafting required gradual elaboration, it offers instant structuring. Where analysis required comparing sources, it simulates perspective-taking.

This cognitive economy has a dual effect. On the one hand, it frees up time and mental energy. The user can concentrate on supervision, validation and orientation tasks. On the other hand, it transforms the very nature of intellectual effort: work is no longer focused solely on raw production, but on the ability to guide and correct.

The analogy with the calculator is illuminating, but insufficient. The calculator automates a deterministic calculation. The AIg produces probabilistic statements. It does not give a single result, but a plausible proposition. The heart of the activity then shifts to the critical evaluation of this proposition.

The cognitive weapon is therefore not so much in the production as in the speed and scale of that production.

The multiplier effect: when mastery amplifies mastery

Any technology benefits more from those who already possess the skills needed to exploit its potential. Generative AI is no exception to this rule.

Those who have mastered argumentative structures, discursive registers and conceptual frameworks will be able to formulate precise, contextualized, iterative queries. They will be able to ask for reformulation, demand clarification of hypotheses, and cross-reference perspectives. The tool thus becomes a gas pedal of intellectual power. It increases productivity without necessarily reducing depth.

Conversely, users with few cognitive tools will tend to accept the first answer as satisfactory. AI then becomes a source of ready-to-use, unquestioned, poorly contextualized statements. What was supposed to be a lever of emancipation can turn into a mechanism of dependency.

In a reading inspired by critical sociology, we could say that cultural capital is now being converted into algorithmic capital. The ability to problematize, to structure and to doubt become the conditions for strategic appropriation of the tool. Inequality doesn't disappear; it changes form.

Invisible asymmetry: from access to orchestration

The most widespread illusion is to confuse access with power. Today, access is relatively democratized: an Internet connection, a user account, and the tool is available. But availability is no guarantee of understanding or mastery.

The real power lies in orchestration. Orchestration means knowing how to set parameters, iterate, compare versions, detect bias, reformulate instructions to refine the response. It means understanding that the system works on the basis of linguistic probabilities, not intrinsic factual verification. It means recognizing that every response is situated, contextualized and dependent on a training corpus.

The asymmetry then becomes subtle. It cannot be seen in the interface, which is identical for all. It lies in the ability to orientate the tool. Two individuals with the same access can produce radically different results in terms of quality, relevance and depth.

A new cognitive divide is thus emerging: no longer just between those who have the information and those who don't, but between those who know how to pilot symbolic devices and those who are subjected to them.

When the tool changes the mind

If generative AI acts as a cognitive weapon, amplifying the capabilities of some while accentuating asymmetries, a more fundamental question emerges: is it also transforming our very way of thinking?

Amplifying the power to act is one thing. Reconfiguring mental structures is quite another. The challenge is no longer limited to the distribution of power, but to the transformation of intellectual operations themselves.

It is this possible mutation between a new alphabet and the externalization of judgment that we must now examine.

New alphabet or externalization of judgment?

If AIg acts as a cognitive weapon, redistributing the power to act, it also operates at a deeper level: that of mental structures. This major technology is transforming valued intellectual tasks and redefining the skills needed to think. Writing enabled stabilized abstraction; printing multiplied voices; the calculator shifted the teaching of mathematics towards modeling rather than manual calculation.

Could AIg represent a new threshold of cognitive transformation? Is it becoming a new alphabet - a medium structuring our relationship to knowledge - or is it encouraging excessive delegation of judgment?

The great cognitive ruptures: what history teaches us

Every technical revolution has given rise to concern: writing would weaken memory, the printing press would dissolve authority, the calculator would destroy mental arithmetic. Yet these tools have gradually been integrated into learning, to the point of becoming invisible.

An alphabet doesn't just add a medium; it reconfigures thinking. It makes analysis, critical distance and cumulative transmission possible. Writing has not eliminated memory: it has transformed it. The calculator didn't abolish mathematics: it shifted the focus to understanding structures.

Generative AI could be part of this continuum. It introduces a new form of linguistic mediation: dialoguing with a machine becomes a structuring skill. Formulating a clear request, contextualizing a problem, refining an instruction: these gestures are perhaps the beginnings of a new literacy.

But the comparison has its limits. Writing and calculators automate specific operations; AI produces plausible statements with no intrinsic guarantee of truth. It doesn't just assist; it simulates intellectual production.

AI as a second writing technology

Generative AI intervenes directly in the act of writing. It suggests formulations, structures arguments and reformulates texts. It acts as a second handwriting, a syntactic mirror capable of imitating discursive codes.

In an educational environment, this ability disrupts traditional reference points. Writing is no longer necessarily the direct trace of individual reasoning. It can become the result of human-machine interaction. Stylistic fluidity no longer guarantees conceptual authenticity.

However, this second handwriting can also become a learning tool. Reformulating a text to improve clarity, comparing several versions of an argument, asking for clarification of hypotheses: these are all uses likely to reinforce metacognitive awareness.

The central question then becomes: does AI serve to produce in place of the learner, or to make the mechanisms of meaning production visible?

The risk of delegation: fluidity without depth

The danger lies not in the tool itself, but in the cognitive economy it creates. When production is instantaneous, the temptation is great to confuse speed with comprehension.

A well-formulated answer can mask a lack of appropriation. A coherent argument may be adopted without having been really constructed. The externalization of writing can gradually lead to the externalization of judgment.

This phenomenon is not necessarily visible. The user may retain the impression of having understood, when in fact he or she has simply validated a plausible formulation. Dependency is not spectacular; it is silent.

The risk is that certain skills will gradually atrophy: editorial stamina, personal formulation, the ability to withstand uncertainty before clarification. It's not technical skills that are being eroded, but intellectual dispositions.

From cognitive transformation to the question of conditions

AIg not only amplifies the power to act, it also intervenes in the very way in which reasoning is elaborated and expressed. It can become a new cognitive alphabet, structuring thought through dialogical interaction. It can also promote an invisible delegation of judgment.

Everything depends on the conditions of use.

If the tool transforms mental structures, then the question is no longer simply one of amplification or delegation. It becomes a question of understanding the mechanisms at work. Who understands how these systems produce their responses? Who can identify their limits, biases and framing effects?

In other words, after amplification and cognitive transformation, we are now faced with an even more decisive question: that of AI literacy as a real condition of power.

Access vs. power: AI literacy as a condition of autonomy

Having considered AIg as a cognitive weapon (amplification) and then as a possible new alphabet (transformation of mental structures), it's clear that the real challenge lies neither in technical access nor in the question of use alone. It lies in understanding.

For the history of technology shows that power does not flow mechanically from the availability of a tool. It depends on the ability to grasp its mechanisms, limits and effects. Generative AI is no exception. It demands a new form of literacy, not just instrumental, but critical.

Three levels of appropriation: from use to awareness

The appropriation of a technical device can be described on three distinct levels.

1. First level: technical access.
   
   Having an account, knowing how to enter a request, getting an answer. This level corresponds to apparent democratization. The interface is simple, the use intuitive. Inclusion seems to be a given.
   
   
2. Second level: operational mastery.
   
   Knowing how to formulate precise instructions, contextualize a problem, iterate, compare versions. Here, the user becomes the pilot. He no longer consumes an answer; he constructs an interaction. The quality of the result depends directly on the quality of the questioning.
   
   
3. Third level: critical understanding.
   
   Understand that the system works through probabilistic language modeling. Identify potential biases in training data. Recognize the epistemological limits [value of knowledge] of a generative model. Distinguish between plausibility and validity.
   
   Real cognitive power lies at this third level. Without it, even sophisticated use remains superficial. AI literacy is therefore not just about "prompting well", but knowing what you're doing when you prompt.

Technical devices and regimes of truth

Every technical device produces a certain regime of truth. It organizes what is visible, dictable and thinkable. By producing fluid, coherent responses, AIg can give the illusion of implicit authority.

Syntactic fluidity thus becomes a factor of credibility. But this credibility is performative: it depends more on form than substance. Users with little training may confuse linguistic coherence with conceptual robustness.

Developing algorithmic literacy means learning to question this regime of truth.

• What implicit assumptions structure the answer?
• What cultural or normative frameworks are reproduced?
• What blind spots remain invisible?
• What uncertainties are masked by the affirmative formulation?

It's not a question of adopting a posture of permanent defiance, but of cultivating an epistemic vigilance. Understanding the system becomes a condition for intellectual autonomy.

Educational responsibility: shaping subjects capable of guiding technology

If power depends on understanding, then the question becomes institutional. Who provides this literacy training? Who explains the mechanisms? Who makes the biases visible?

Leaving learners to face the tool alone would be tantamount to naturalizing its effects. Conversely, explicitly integrating AI into curricula enables spontaneous use to be transformed into reflective learning.

Training in AI literacy involves a number of shifts:

• Teaching the distinction between assisted and autonomous production.
• Evaluating not only the result, but also the process of interaction with the tool.
• Explaining the technical and ethical limits of generative models.
• Encourage critical reformulation rather than immediate acceptance.

Responsibility doesn't stop at preventing aberrations; it means directing use towards explicit educational goals. The challenge is not to restrict access, but to enrich understanding.

From individual literacy to the institutional commons

Generative AI reveals a fundamental distinction: access is only a minimal condition; power lies in the ability to understand and guide.

Without critical literacy, the tool risks accentuating the asymmetries already described. With it, it can become a lever for autonomy and discernment. But this literacy cannot rely solely on individuals. It presupposes pedagogical frameworks, institutional choices and explicit governance.

In other words, having examined amplification and cognitive transformation, we must now broaden our thinking: how can we transform this concentrated technology into a truly shared resource?

The question of power then becomes a question of commonality.

Towards an educational commons: governing, sharing, instituting

If generative AI constitutes a technology of cognitive power, if it transforms our mental structures and demands a specific literacy, then the ultimate question is neither technical nor individual. It is political and institutional.

Such a structuring technology can remain concentrated in the hands of a few industrial players, a few experts capable of piloting it, or it can become a shared resource, integrated into a common culture. The transition from the cognitive weapon to the educational commons presupposes specific conditions: pedagogical, ethical and organizational.

Establishing explicit pedagogical conditions

Transforming AI into a shared resource means making it an object of instruction, not just a peripheral tool. Implicit integration encourages spontaneous use; explicit integration builds skills.

Several principles can structure this institutionalization:

• Alternate between assisted and independent production.

The tool must not systematically replace intellectual effort. It must be used at identified moments, with clear objectives.

• Evaluate processes as well as results.

Documenting interaction with AI, making reformulation choices explicit, justifying modifications made: these practices shift the center of gravity towards reflection.

• Make limits visible.

Integrate activities in which biases, errors or approximations are analyzed collectively. The tool thus becomes a support for critical discussion.

It's not a question of framing to restrict, but of framing to make intelligible. Learning AI must be as explicit as learning to write or to reason scientifically.

Governance and sovereignty: an institutional responsibility

Shared transformation is not just a matter of pedagogical practices. It requires clear governance.

Educational institutions need to address a number of issues:

• Choice of tools and technological dependence.

What models are used? Under what conditions of confidentiality and data processing?

• Transparency and traceability.

Are the rules of use explicit? Are responsibilities identified?

• Training the trainers.

Collective appropriation presupposes that teachers themselves have sufficient critical understanding.

At stake is cognitive sovereignty. An institution that unthinkingly delegates its pedagogical systems to systems it doesn't understand is partially abandoning its mission of autonomous training. Conversely, an institution that structures usage transforms potential dependence into collective capacity.

From user-friendliness to collective intelligence

For a technology to become a common asset, it must be appropriable, discussable and transformable. A user-friendly tool is not a simplified tool; it's a tool that increases the ability to act without alienating.

AIg can contribute to enhanced collective intelligence:

• co-construction of syntheses,
• confrontation of arguments,
• rapid exploration of hypotheses,
• diversification of perspectives.

But this collective intelligence does not emerge spontaneously. It presupposes a culture of debate, an ethic of responsibility, and an awareness of technical limitations.

A commons is not simply shared access; it is a resource governed collectively according to explicit rules. Without governance, industrial and cognitive concentration impose themselves. With it, the tool can become a means of emancipation.

From concentrated power to shared responsibility

AI is neither intrinsically emancipatory nor inevitably unequal. It amplifies existing capabilities, reconfigures mental structures and demands a new literacy.

The decisive question is how to institutionalize it. Will it be left to market dynamics and individual appropriation, or enshrined in an explicit educational project?

Transforming a cognitive weapon into a shared alphabet requires more than access: it requires frameworks, training, governance and assumed responsibility.

And then there's the ultimate question, which goes beyond the tool itself: are we ready to consider the mastery of cognitive technologies as an essential component of the educational common good?

Only then will it be possible for AIg to become something other than a factor of asymmetry, and become a lasting part of a culture of intellectual autonomy.

Power as collective learning

Generative AI presents us with a structuring ambiguity. It amplifies the power to act, while making gaps in control more visible. It can become a gas pedal of intellectual emancipation or a subtle factor of cognitive asymmetry. It can support learning or silently externalize judgment.

It all depends on the conditions under which it is used.

The mistake would be to reduce the debate to a simplistic alternative between enthusiastic adoption and defensive refusal. What's at stake is not usage per se, but the ability to shape that usage. Technical access is only a minimum threshold; the real power lies in understanding the mechanisms, in the ability to question the answers produced, in the ability to articulate assistance and autonomy.

AIg is not just another teaching tool. It goes to the heart of symbolic activity: writing, arguing, synthesizing, problematizing. As such, it redefines the skills needed to learn and teach. To ignore this mutation would be to allow a silent cognitive divide to develop.

Conversely, explicitly instituting it in curricula - by developing a demanding algorithmic literacy, training teachers, clarifying usage frameworks and responsibilities - would make it possible to transform a concentrated technology into a shared resource.

The central question, then, is not "Should we use generative AI?"; it is more fundamental: "Do we want to train assisted users or subjects capable of guiding the technologies that structure their cognitive environment?"

If schools take on this role, AI could become a new alphabet, not a substitute for reasoning, but a support for deepening it. If it does not, power will remain concentrated between those who master the devices and those who are subjected to them.

IAg thus reveals a broader truth: technological power is never given, it is learned. And it is in this collective learning that the possibility of a true cognitive commons is at stake.

Illustration: AIg, the new cognitive weapon
Generated by AI (Canva) - Flavien Albarras

References

Auroux, S. (1994). La révolution technologique de la grammatisation: Introduction à l'histoire des sciences du langage. Mardaga.

Digital Promise. (2024). AI literacy: A framework to understand, evaluate, and use emerging technology. Washington, DC: Digital Promise. https://digitalpromise.org/2024/06/18/ai-literacy-a-framework-to-understand-evaluate-and-use-emerging-technology/

Florey, S. (2023). Reconfiguration du littéraire par le numérique: Quelles potentialités didactiques? Revue de recherches en littératie médiatique multimodale, 18, 39-55. https://doi.org/10.7202/1108693ar

Fricker, M. (2007). Epistemic injustice: Power and the ethics of knowing.

Kvasny, L. (2005). The Role of the Habitus in Shaping Discourses about the Digital Divide. J. Computer-Mediated Communication, 10. https://doi.org/10.1111/j.1083-6101.2005.tb00242.x

Long, D., & Magerko, B. (2020). What is AI Literacy? Competencies and Design Considerations. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, CHI '20, 1-16. https://doi.org/10.1145/3313831.3376727

Matjie, M. A., Nethavhani, A., & Matlakala, M. (2026). AI and the digital divide in education. Frontiers in Computer Science, 8. https://doi.org/10.3389/fcomp.2026.1759027

OECD. (2022). OECD Digital Education Outlook 2021: Pushing the boundaries with AI, blockchain and robots. OECD Digital Education Outlook, 2021. https://doi.org/10.1787/d5fe6bd0-fr

Ong, W. J. (1986). Writing is a technology that restructures thought. In G. Baumann (Ed.), The written word: Literacy in transition (pp. 23-50). Oxford, United Kingdom: Clarendon Press. PDF version: https: //worrydream.com/refs/Ong_1985_-_Writing_is_a_Technology_that_Restructures_Thought.pdf

UNESCO. (n. d.). AI and education: Guidance for policy-makers-UNESCO Digital Library. Retrieved March 8, 2026, from https://unesdoc.unesco.org/ark:/48223/pf0000376709

UNESCO. (2023). Guidance for generative AI in education and research. Paris, France: UNESCO. French version: https: //www.unesco.org/fr/digital-education/artificial-intelligence

Yang, Y., Zhang, Y., Sun, D., He, W., & Wei, Y. (2025). Navigating the landscape of AI literacy education: Insights from a decade of research (2014-2024). Humanities and Social Sciences Communications, 12(1), 374. https://doi.org/10.1057/s41599-025-04583-8