Alors que l’intelligence artificielle générative produit des textes, des images et des raisonnements toujours plus convaincants, une question éducative essentielle émerge : que signifie encore comprendre ? À l’ère de la surcharge informationnelle, former une « tête bien faite » ne consiste plus à accumuler des savoirs, mais à apprendre à juger de leur profondeur, de leur validité et de leur inscription dans le réel.

Des étudiants peuvent rendre aujourd’hui des travaux irréprochables sur la forme : structurés, argumentés, parfois brillants. Pourtant, lorsqu’on les interroge, un malaise affleure. Ils peinent à expliquer ce qu’ils ont réellement compris, à justifier leurs choix, à relier ce qu’ils ont produit à une expérience vécue ou à une situation concrète. L’intelligence artificielle (IA) générative n’est pas toujours la cause directe de cette situation, mais elle en est un révélateur puissant. Car si produire de l’information n’a jamais été aussi simple, comprendre ce que l’on fait n’a jamais été aussi exigeant.

Connaître, savoir, comprendre : une distinction devenue centrale

À l’ère de l’IA, la question éducative ne peut plus être pensée en termes d’accumulation de connaissances. Elle impose de clarifier ce que l’on entend par connaître, savoir et comprendre, et d’interroger la manière dont ces dimensions s’articulent dans les processus d’apprentissage.

Deux grandes traditions épistémologiques permettent d’éclairer cette distinction. Le scientifique Michael Polanyi a montré que toute connaissance humaine comporte une part irréductiblement tacite : elle est enracinée dans l’expérience, l’action et l’engagement du sujet. « Nous savons plus que ce que nous pouvons dire », écrit-il, soulignant que la compréhension précède souvent sa formulation explicite. Cette connaissance en acte, souvent implicite, se construit dans le faire, l’essai, l’erreur et la confrontation au réel.

À l’inverse, le philosophe Gaston Bachelard a établi que le savoir scientifique ne procède pas d’un simple prolongement de l’expérience. Il exige une rupture avec les évidences premières et avec l’opinion, au prix d’un travail de construction rationnelle, critique et abstraite. « La science ne procède pas de l’opinion », rappelait-il, insistant sur la nécessité de former l’esprit à poser des problèmes plutôt qu’à accumuler des réponses.

Former une « tête bien faite », ce n’est donc ni accumuler des savoirs abstraits, ni se contenter d’une expérience brute. C’est apprendre à tenir ensemble ces deux dimensions : l’expérience vécue et la construction conceptuelle, l’action et la réflexivité.

Ce que l’IA sait faire – et ce qu’elle ne peut pas faire

Les systèmes d’intelligence artificielle excellent précisément là où le savoir est formalisable : calcul, synthèse, reproduction, mise en forme. Ils prennent en charge une part croissante du savoir explicite, stabilisé, calculable. Mais ils opèrent dans un régime spécifique : celui de la corrélation statistique et de la production d’énoncés plausibles.

L’IA ne connaît pas le monde, elle ne le comprend pas. Elle n’a ni expérience, ni rapport incarné au réel, ni accès aux conditions de pluralisme des phénomènes qu’elle décrit. L’information qu’elle génère est par nature probabiliste – elle repose sur des calculs de vraisemblance issus de corrélations statistiques plutôt que sur une compréhension des causes, contingente – elle dépend des données, des contextes d’énonciation et des paramètres techniques, et révisables – en ce sens qu’elle peut être corrigée, contredite ou reformulée à tout moment sans que cela n’implique une progression interne de la compréhension.

Cette distinction est aujourd’hui au cœur des travaux contemporains sur les usages éducatifs de l’IA, qui montrent que l’automatisation de certaines tâches cognitives peut, si elle est mal encadrée, appauvrir l’exercice du jugement critique.

Plus les productions de l’IA deviennent convaincantes, plus le risque est grand de confondre cohérence formelle et compréhension réelle, à savoir paraître véridique en lieu et place d’un énoncé prudent et ouvrant le dialogue.

Mesurer la complexité : une compétence qui s’apprend

Face à cette situation, un enjeu éducatif majeur émerge : la capacité à mesurer la complexité des choses. Distinguer ce qui relève de la surface informationnelle de ce qui engage une compréhension structurée. Apprécier les niveaux de profondeur d’un problème, d’un système ou d’une situation.

Or cette capacité ne se décrète pas. Elle se construit progressivement par l’expérience du réel. Elle suppose un travail actif de confrontation entre ce que l’on anticipe théoriquement et ce que révèle l’épreuve de la réalisation concrète. C’est dans l’écart, toujours instructif, entre le modèle et l’expérience que s’affinent les critères de jugement et que se développe une intelligence véritablement située, au sens où elle articule savoir formalisé et connaissance vécue.

Le savoir ne devient opérant qu’à condition d’être éprouvé, mis en tension avec le réel, réajusté à la lumière de ses résistances et de ses surprises. Inversement, l’expérience brute, si elle n’est pas reprise dans un cadre réflexif et conceptuel, demeure muette et difficilement transmissible. La formation doit donc organiser les conditions de cette circulation exigeante entre théorie et pratique, abstraction et incarnation.

Une révolution pédagogique autant que technologique

L’intelligence artificielle ne transforme pas seulement nos outils. Elle intervient au cœur même des fonctions cognitives supérieures : mémoire externalisée, accès instantané à l’information, génération de raisonnements apparents. Là où les technologies précédentes amplifiaient des capacités humaines déjà constituées, l’IA en reconfigure désormais l’équilibre.

L’enjeu éducatif se déplace en conséquence. Il ne s’agit plus principalement d’apprendre à produire ou à restituer de l’information mais d’apprendre à en évaluer la profondeur, la cohérence, les conditions de validité et les effets dans le réel. Cette mutation rejoint les analyses du sociologue Edgar Morin sur la pensée complexe, qui soulignent la nécessité de former des esprits capables de relier, de contextualiser et d’affronter l’incertitude plutôt que de réduire le réel à des réponses simplifiées.

Des travaux récents en sciences cognitives et en sciences de l’éducation montrent d’ailleurs que l’usage substitutif de l’IA peut conduire à une forme de délégation cognitive excessive, réduisant l’engagement intellectuel et la mémorisation à long terme, là où un usage réflexif et critique peut au contraire renforcer l’apprentissage.

Former des ingénieurs – et des citoyens – capables de juger

Former une tête bien faite à l’ère de l’IA implique ainsi de ne pas confondre délégation cognitive et renoncement intellectuel. Il s’agit de former des sujets capables de faire usage de systèmes puissants sans s’y soumettre, capables de maintenir une exigence de sens là où la machine ne produit que de la forme.

Chez IONIS, le développement des IONIS Institute of Technology (I2T) sur nos campus procède d’une conviction forte : si nos étudiants ingénieurs doivent maîtriser les technologies de l’intelligence artificielle, ils doivent tout autant apprendre à en éprouver les limites par la confrontation au réel. Le laboratoire, l’atelier et l’expérimentation deviennent alors des lieux centraux de formation du jugement.

Former de bons ingénieurs – et plus largement des citoyens éclairés – consiste à cultiver un esprit critique, mesuré et évolutif, nourri par l’expérience concrète, le faire et le défaire. À l’ère de l’IA, la question essentielle n’est donc pas seulement ce que nous attendons de la machine mais bien ce que nous attendons de l’humain : sa capacité à comprendre, à créer et à décider avec discernement dans des environnements incertains et technologiquement augmentés.

Clément Duhart ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d'une organisation qui pourrait tirer profit de cet article, et n'a déclaré aucune autre affiliation que son organisme de recherche.

The situation in the Middle East represents a negative framework for Italy, which, like its European partners, is a mere spectator watching a conflict that it does not endorse.

Italy is both an exporting nation, sensitive to global economic cycles, and a country characterised by its dependence on fossil fuels, with Italian electricity production largely powered by natural gas.

Both the rise in gas prices and a slowdown in global trade pose a dangerous economic pincer for Italy. This widely shared assessment within the Italian political class underpins Italy’s critical stance towards the conflict.

EU alignment versus handling US relations with care

It is also important to note a shift in Italy’s positioning. Since the beginning of Trump’s presidency, Giorgia Meloni’s government has shown conciliation towards Trump.

While Italy maintained a form of European orthodoxy – supporting Ukraine, expressing a favourable position towards Denmark regarding Greenland – it also tried to please both sides, paranoidly avoiding initiatives that could be perceived as distancing itself from Washington, especially due to concerns over tariffs. From this perspective, Italy remained in the background regarding European defence initiatives led by France and the United Kingdom, from the Coalition of the Willing to the new concept of “Forward Deterrence,” which sees France collaborating with major European countries, including Germany.

In some ways, this Italian caution may have contributed to a misjudgment by the Italian government, which, for example, did not anticipate Germany’s growing distance from the United States.

In February, Italy’s participation as an observer in Trump’s “Board of Peace” had already raised many questions.

This choice by Meloni’s government appeared as a continuation of an open line towards the American administration, a traditional position she has repeatedly voiced when evoking her vision of Western unity, conceived as a fundamental project. At the time, the left-wing opposition expressed its discontent, while even the Vatican showed its perplexity.

Italy seeks distance on Iran

Following Israel’s and the United States’ military action against Iran, the Italian Prime Minister expressed euphemistic distancing from the United States, declaring in the Chamber of Deputies that the actions of the United States and Israel were taking place outside international law. This amounted to an implicit rejection by Italy, which remains constitutionally committed to legality. Although not a direct condemnation of the Trump administration, this moment marked a turning point, breaking with a government that had previously appeared consistently lenient towards Trump’s positions.

This shift was later confirmed by Italy’s signing, alongside Canada, France, Germany and the United Kingdom, of a letter refusing to send military capabilities to the Strait of Hormuz following a request by Donald Trump.

Keeping energy options open with Tehran

Further factors help explain this position. Italy has always cultivated its relationship with Tehran. Iran had been a privileged partner in Italy’s energy policy, established by ENI under Enrico Mattei since 1957, an agreement that, in some ways, endured even after the 1979 revolution.

Despite sanctions, Italian diplomacy has always sought to maintain channels of dialogue with Tehran. On the other hand, Italians have always been, rightly, extremely critical of military ventures aimed at regime change. In 2011, they had warned – though unheeded – that the fall of Gaddafi’s regime would lead to a scenario of great instability. They were also disillusioned by the withdrawal from Afghanistan, a theatre in which they had invested heavily and where they perceived a form of American betrayal.

Generally, Italy, a country marked by pacifism, does not believe in solutions imposed by force. Thus, the March 2026 attack represents a worst-case scenario for Italians: it creates a risk for their economy, marks a rupture with a country long considered a partner, and defines a military spiral that Italians know to be harmful. This conflict, along with pressures at home, has put the Meloni government in a state of crisis, as it has made Atlantic loyalty a hallmark, in keeping with the line expressed by the Italian right since Silvio Berlusconi.

The Italian context

Italian domestic politics have been characterised by competition over the recent government-backed justice referendum – a key test for Meloni, who, despite this defeat, is continuing as PM. The events in Minneapolis and the blatant violations of freedoms by U.S. federal forces have contributed to Italy’s cautious stance towards judicial reform, perceived as an untouchable shield for democracy.

References to Trump now act as a repellent for the entire Italian political class, including the right. This factor played a role in the vote against the referendum, which is a major blow for Meloni. It is also worth noting that the visit of Palantir CEO Peter Thiel to Rome in March 2026 sparked both calculated political indifference and an outcry among Catholic circles: Thiel was labelled a heretic and an opponent of the social doctrine of a religious institution that remains steadfast in its defence of democracy.

This distance from America’s tech-right contrasts with the warm welcome Giorgia Meloni had reserved for Elon Musk until 2024.

The radicalisation of Trump’s position is pushing the Italian government to clarify its pro-European choice, a painful evolution for a political camp whose visceral Atlanticism feels betrayed by the United States.

A weekly e-mail in English featuring expertise from scholars and researchers. It provides an introduction to the diversity of research coming out of the continent and considers some of the key issues facing European countries. Get the newsletter!

Jean-Pierre Darnis ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d'une organisation qui pourrait tirer profit de cet article, et n'a déclaré aucune autre affiliation que son organisme de recherche.

In the wake of the final round of France’s local elections, we asked political pundit Frédéric Sawicki what the mayoral vote tells us about the political future of the two main contenders – Rassemblement National (RN) and La France Insoumise (LFI)?

The Conversation: What was the outcome of France’s 2026 mayoral elections?

Frédéric Sawicki: Let’s start by recalling that, in 9 out of 10 municipalities, the elections were conclusive right from the first round. The second round mainly concerned urban France. In one in two cities, at least three lists were running against each other, confirming the tripolarisation of the political arena. Overall, political stability dominates because, unlike in 2008 or 2014, these mayoral elections were not an opportunity to sanction a majority or a government, due to a massive lack of lists of candidates representing the “presidential majority.”

Consequently, the few shifts from one camp to another are often explained by local concerns. If the Parti Socialiste (PS), France’s centre-left establishment, wins Saint-Étienne, it may well have something to do with the sex tape scandal surrounding the outgoing mayor. And vice versa in Brest in Brittany where the PS lost, which the ailing reputation of the outgoing mayor who had been in office since 2001, no doubt had a hand in. There are multiple cases of this.

We also note a certain number of “sociological constants.” France’s largest cities always vote massively to the left. In cities with tourist economies, a strong presence of independent professions, small employers, retailers, military-related populations (in Toulon, for example), or well-off retirees (Nice, Menton or Cagnes-sur-Mer), the vote is strongly right or far-right leaning.

On the whole, we noticed as many swings from the right to the left and vice versa, as movements within each camp: socialist or communist cities became LFI territory (Saint-Denis, Vénissieux), “green party” cities shifted towards the centre left (Strasbourg or Poitiers). On the other end of the spectrum, some right-wing or centre right cities shifted to the far right or to its allies, for example, as was the case in Nice with Eric Ciotti.

The slides from left to right were few and far between and were balanced with slides in the opposite direction. Over the municipalities of more than 50,000 inhabitants, the left and the ecologists lost Bordeaux, Clermont-Ferrand, Brest, Avignon, Besançon, Poitiers, Cherbourg, but won Saint-Étienne, Nîmes, Amiens, Roubaix, Aubervilliers and Pau.

RN’s wins in smaller towns were quite a result…

F.S.: One of the new phenomena to come out of the vote is the solid bolstering of far right party Rassemblement National’s “mayoral anchorage”. In 2020, the RN claimed only 17 cities, including seven with more than 20,000 inhabitants. This time, it won 60. Toulon or Nîmes were, without doubt, huge losses for RN, but they managed to win over many medium-sized cities, such as Carpentras, La Seyne-sur-Mer, La Flèche, Vierzon, Montargis, Liévin… RN strongholds are primarily located in the former mining Pas-de-Calais area, the Côte d'Azur, and also in Occitania (Agde, Castres, Carcassonne, Montauban).

The RN has particularly anchored its presence in the cities it conquered in 2014 and 2020; strengthening its credibility is essential for the RN ahead of the 2027 presidential elections.

Did LFI – the radical left party come out of these local elections any stronger?

F.S: LFI’s breakthrough became apparent as it won over several cities with more than 30,000 inhabitants – Saint-Denis, Roubaix, Saint-Paul and Le Tampon (France’s overseas territory – Island of La Réunion), Vénissieux, Vaux-en-Velin, La Courneuve, Creil, which is far from insignificant considering they started from scratch. However, this breakthrough must be put into context.

France’s radical left suffered losses in situations where they had taken the lead from the left (in Toulouse or Limoges). The party led by Jean-Luc Mélenchon failed to reclaim towns from the right, with the exception of Roubaix.

LFI specifically poised itself as an essential partner, with the threat of bringing down the centre-left PS if no alliance was formed. However, when the socialists and allies refused the merger, they were rarely defeated. LFI therefore failed to demonstrate its ability to hit hard – the example of Paris was emblematic, but Rennes and Montpellier are other fine examples.

Finally, in the context of alliances with the environmentalists, in particular, we note LFI losses in cities such as Strasbourg, Poitiers, Besançon. When the alliances made it possible to win (Grenoble, Lyon, Tours), LFI will not be represented at executive level because the mergers were only a “technicality.” The LFI representatives therefore remain confined to the role of opponent, which doesn’t give the party the ability to govern and appearing as a party “purely of the opposition” poses a real problem for obtaining supreme (presidential) power in 2027.

We noticed that the centre and the right strengthened their alliance during the mayoral runoffs…

F.S.: There were a lot of lists where the centre right, Horizons, MoDem and Renaissance, joint forces with conservative candidates, Union des démocrates et indépendants (UDI) or Les Républicains (LR), from the first round. In the second round, many alliances were formed. This municipal election is therefore part of the continuity of the parliamentary alliance between the central block and the right since 2022 and especially since 2024, and foreshadows what could happen in the presidential election. After the second round, we heard calls for LR, Horizons and the Macronists to have a common arrangement and one candidate. With his win in Le Havre, Édouard Philippe delivered a speech aimed at federating these forces – even though the battle remains very open.

There has been a lot of talk about “infighting within the left-wing,” with a clear divide between part of the socialists and LFI. What are the stakes for 2027?

F.S.: The fracture between those who refuse any alliance with LFI and the others has been reinforced with this vote. But what we notice is that the autonomous centre hardly exists any more, the central block is now clearly uniting with the right. So what is the alliance strategy of those who, like Raphaël Glucksmann or François Hollande, advocate a rapprochement with the centralists? If the Socialist Party comes out stronger after this election, it’s because it joined forces from the first round, with the ecologists, the communists, Place publique and former LFI supporters. The only path that seems to be emerging for the left is to work on this foundation by making it stronger – without LFI – while showing LFI voters that they are not “inaudible” and that their aspirations are heard.

Some observers feared a “fusion” between the right and the far right. While this didn’t happen at party level, we have, on the other hand, witnessed moderate right-wing and far right-wing voters becoming closer…

F.S.: Local initiatives to bring the RN closer to conservative candidates were immediately condemned by LR. The fact that Bruno Retailleau did not call for Christian Estrosi to vote against Éric Ciotti in Nice caused outrage. On the other hand, voters’ choices at the ballot showed substantial porosity between the right and the far right. The presence of LFI candidates acting as a foil, RN voters voted LR to block the left, and LR voters voted RN for the same reason. It’s clear in Brest, Clermont-Ferrand, Limoges, Besançon, and Marseille that conservative LR Martine Vassal’s electorate (LR) was halved between the first and second rounds, in favour of Franck Allisio (RN).

Finally, who came out as the winner? Does this mean RN might be heading for success in 2027?

F.S.: The RN enjoyed a positive dynamic, strengthening its local presence is an important step in the right direction ahead of the senatorial elections, but also presidential and next legislative elections. That said, the FN remains on the outside looking in with regards to France’s largest cities, which is a major weakness for setting its sights on the Élysée palace. How could a party claim power at the national level while being absent in the cities where most of France’s intellectual and economic activity is concentrated? The RN represents around 30% of voters: it can win in 2027 against an isolated Jean-Luc Mélenchon, or if the left (excluding LFI) and the right remain divided. The potential success of the RN in 2027 is therefore not inevitable provided that the left or the right come together. This is what François Mitterrand managed to do in 1981 and 1988. That’s what Jacques Chirac did in 1995, like Nicolas Sarkozy in 2007. It is now up to the political forces to play… their cards right.

Interview by David Bornstein, Political Editor at The Conversation France.

A weekly e-mail in English featuring expertise from scholars and researchers. It provides an introduction to the diversity of research coming out of the continent and considers some of the key issues facing European countries. Get the newsletter!

Frédéric Sawicki ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d'une organisation qui pourrait tirer profit de cet article, et n'a déclaré aucune autre affiliation que son organisme de recherche.