The New Lucid Ignorance: When AI Gives Us the Illusion of Thinking

There's a paradox lurking among us, silent and glittering: assisted ignorance . It's no longer the old presumption of the amateur who thinks he's a master. It's a new, lucid ignorance, which dresses itself up as competence because an algorithm has lent it the right words. The Dunning-Kruger effect , that old ghost that told us that the least capable overestimate themselves and the most competent underestimate themselves, has now found a powerful ally: artificial intelligence. No longer just a cognitive bias, but a democratization of illusion . All it takes is a prompt, and anyone can feel like an expert.

Scientists at Aalto University demonstrated this with cruel clarity: 500 people, 20 logic problems . With ChatGPT , the results improve. But self-perception collapses. Those who use AI feel brilliant, even when they make mistakes. And the more you know the machine, the more you convince yourself of your infallibility. It's the mirror oracle syndrome : the machine gives you answers, you look in the mirror and believe yourself to be wise.

Delegated thinking

Most participants asked the chatbot a single question, without verifying anything. This is " cognitive offloading ": thoughts are offloaded, like a useless burden, entrusted to an external engine. It's no longer the effort of reasoning, but laziness disguised as efficiency AI psychosis is born : not a technological delirium, but a slow detachment from critical reality.

Automatic flattery

Machines don't contradict you. They caress you. They make you feel brilliant. It's " sycophancy ": algorithmic flattery . A digital chorus that always tells you " you're right ," even when you're wrong. And you, lulled by this litany, stop doubting. But without doubt, there is no thought. Without friction, there is no truth. Without tension, even less.

The new ignorance

AI doesn't make us wiser. It makes us safer. And safety without awareness is the new ignorance . An ignorance that doesn't recognize itself as such, because it wears the mask of expertise . It's the economy of indistinguishability applied to the mind: when everyone seems brilliant, no one really is . How then can we distinguish? How can we discern? How can we distinguish when AI has made signals indistinguishable? It's no longer enough to read a well-written text or an impeccable CV: those are empty shells, infinitely replicable. Discernment today must shift to other levels, more difficult to simulate. We'll get to that shortly...

The task that remains

The problem isn't the machine. It's how we use it. Delegating thought means stopping thinking. The real challenge isn't learning to ask a chatbot questions better , but learning to doubt what it gives us back . Solving problems isn't enough: we must recognize its limitations, cultivate uncertainty, and defend the right to doubt.

I read this article a while back in The Atlantic and I've been stuck in my head ever since. I don't know about you, but during our recent hiring process, we've noticed something strange. The applications are all perfect, too perfect. All mention every single requirement in the ad. All demonstrate "genuine passion" for the role. An article in The Economist talked about it a few days ago. Before ChatGPT, cover letters were an important signal for companies. Writing a decent one took time, care, and cognitive effort. Those who wrote them well showed they truly cared about the position. It was a sort of imperfect but functional filter. Now? Now anyone can produce a perfectly targeted cover letter in 30 seconds. Hundreds of times a day. The result? Many companies have stopped reading them. Because when everyone seems like an excellent candidate on paper, no one actually is. The signal has collapsed. A study analyzes Freelancer(.com), pre- and post-ChatGPT. Before AI, a well-written cover letter was worth $26 more per task. After AI: that premium value has disappeared. The average quality of selected candidates has deteriorated. The risk is that strong workers will no longer be able to stand out, while weak workers send out 500 applications a day with AI. Thus, according to The Economist, companies lower salaries because they no longer know who they're hiring. They hire worse, more often. It's the economy of indistinguishability. When everyone can appear excellent. AI was supposed to democratize access to opportunity, and in a certain sense it has done so. But by democratizing the signal, it has destroyed it. So the question becomes: in a world where anyone can appear competent... how do you recognize who truly is? The answer we've given ourselves at Marketers is simple: go back to the things AI can't fake. We've increased open-ended questions, making them more personal, intimate, and narrative. Questions that only a human can address. And we've introduced practical exercises on real-world cases, which require intuition, empathy, the ability to read others and make non-obvious and difficult decisions. Because you can generate perfect text with a language model, but you can't simulate responsibility when faced with difficult choices.

Dario Vignali

AI has rendered traditional signals (cover letters, "perfect" CVs) indistinguishable, as you mentioned, and so we need to invent new filters that aren't easily replicated by a linguistic model. There are various solutions. Realistic case studies: simulations of real-world business problems, where the candidate must propose practical solutions. Paid micro-tasks: small, real-world tasks, even just a few hours long, that demonstrate the quality of the work and how to address unexpected events. Decision-making tests: ambiguous scenarios where there's no "perfect" answer, but where the ability to reason and choose is assessed. Narrative interviews: questions that require personal stories ("Tell me about a failure and how you handled it"), difficult to simulate with generated text. Evaluating empathy and collaboration: group exercises, role plays, or team dynamics that demonstrate how the candidate interacts. Observable soft skills: listening skills, adaptability, time management, and interpersonal skills. If AI has "killed" "superficial signals," then the real discriminator becomes what cannot be automated, and empathy is one of these elements.

Three levels of discernment

1. The level of living gesture

   • Not the text produced, but the way it is produced.

   • Ask the candidate to think on the fly , to argue without having to "consult" a digital oracle.

   • Here we see the difference between someone who has internalized knowledge and someone who is merely miming it.

2. The level of the relationship

   • Empathy, the ability to listen, to modulate language based on the other person.

   • AI can generate words, but it cannot sense the tension of silence, nor read the fragility of an interlocutor.

   • Distinguishing means observing how a person moves in human space, not just in linguistic space.

3. The level of responsibility

   • Give ambiguous cases, without a perfect solution.

   • Evaluate not the "right" answer, but the courage to make a decision and stick with it.

   • AI can offer infinite options, but it cannot carry the burden of a choice.

The new criterion

Discernment today means shifting the filter from product to process . Don't ask " what you can write ," but " how you think, how you listen, how you decide ." It's a radical shift: from documents to presence, from appearances to responsibility.

A synthetic formula

• Perfect text → collapsed signal.

• Imperfect process → authentic signal.

True talent is recognized not in replicable perfection, but in unrepeatable imperfection: that which arises from the human being who thinks, makes mistakes, corrects, listens, decides..

AI has gifted us with an extraordinary power: the speed of externalized thought. But if we don't learn to curb our egos, inflated by digital flattery, we risk becoming lucid idiots : convinced we know, incapable of thinking . True intelligence today is neither artificial nor natural . It is critical . And without criticism, even the most brilliant mind becomes an empty simulacrum.