🔗 Share this article

This past summer, I found myself navigating in congested traffic on the scorching streets of Marseille. At a crossing, my friend in the front seat advised a right turn toward a renowned spot for bouillabaisse. However, the navigation app on my phone commanded us to go forward. Fatigued and in a stifling car, I heeded the algorithm's advice. Minutes later, we were stuck at a roadwork site.

A minor event, maybe. But one that captures a central dilemma of our era, where technology touches nearly every facet of existence: who gets our trust to a greater degree – fellow humans and our own intuition, or the machine?

The Enlightenment's Promise and Our Modern Relapse

The renowned German philosopher Immanuel Kant once defined the Enlightenment as "humanity's emergence from its self-imposed nonage." This state, he argued, "is the inability to use one's own understanding without direction from an external source." For centuries, that directing force for human thought was frequently the clergyman, the monarch, or the feudal lord – entities claiming to speak for divine will. To explain phenomena like changing seasons, people sought answers in religion. In organizing the social world, from economics to matters of the heart, religious doctrine served as the main compass.

“Sapere aude!” or “Dare to use your reason!”

Kant maintained that humans always possessed the ability to think rationally. They simply lacked the confidence to use it. With revolutions in the 18th century, a fresh era emerged: logic would replace blind faith, and the intellect, freed from dogma, would become the driver of advancement and a better world.

Now, 250 years later, one might wonder if we are slipping back into a form of dependency. An app dictating a driving route is merely the start. Artificial intelligence risks becoming our new "other" – a unseen guide that influences our decisions and actions. We are in danger of ceding the hard-won courage to think independently – and this time, not to gods or kings, but to computer programs.

The Rapid Rise and Subtle Dangers of AI Dependence

ChatGPT debuted only in 2022, and already a recent study found that an vast number of respondents had used AI in the previous six months. Whether contemplating a breakup or selecting a vote, individuals are increasingly turning to algorithms for counsel. Data suggests a large majority of user queries concern personal life matters. Even more striking than our reliance on AI for judgment is what occurs when we let it express for us. Composition is now one of the top uses for generative AI, just behind practical requests. The late American author Joan Didion once remarked, “I write purely to discover what I am thinking.” What occurs when we stop composing? Do we cease discovering?

Worryingly, emerging research suggests the answer may be affirmative. A study from the Massachusetts Institute of Technology used brain monitoring to observe the cognitive activity of participants who had could use AI, traditional search engines, or no aids. Those who could rely on AI showed the lowest brain activity and had trouble accurately recalling their own work. Perhaps most concerning was that after a few months, participants in the AI group became increasingly reliant, pasting large sections of text.

“Laziness and cowardice,” Kant wrote, “are the reasons why so great a proportion of men … stay in perpetual nonage.”

Certainly, AI's attraction lies in its convenience. It is fast, minimizes work and – importantly – offers a new method to offload accountability. In his 1941 book, Escape from Freedom, the German psychoanalyst Erich Fromm proposed that the appeal of authoritarianism could be partly explained by a human tendency to surrender autonomy in exchange for the comforting security of obedience. AI presents a digital avenue for surrendering the weight of having to decide for oneself.

The Black Box Problem: Trust Without Understanding

AI's primary draw is its capacity to perform tasks outside human capability – analyzing vast datasets at lightning pace. Sitting in the car in Marseille, this was, after all, why I chose to trust the machine over my companion (a decision she interpreted as an affront). With knowledge of all the data, certainly the app had superior insight – or so I believed.

The fundamental problem is that AI operates as a opaque system. It produces knowledge, but not always fostering human comprehension. We do not truly know the process behind its conclusions – even its programmers acknowledge the opacity. Nor can we verify its reasoning against transparent standards. So when we heed AI's recommendation, we are not being guided by reason. We are returning to the domain of belief. In dubio pro machina: when in doubt, trust the machine – that may become our future credo.

Using Without Losing: The Essential Challenge

AI can be a powerful tool for humanity in rational inquiry. It can help inventing drugs, free us from "bullshit jobs", or manage taxes – work that are repetitive and are unfulfilling. This is beneficial. But Kant and his contemporaries did not advocate for enlightenment just so humans could assemble better furniture or have more leisure. Independent thought was not merely about efficiency – it was a practice of liberty and human emancipation.

Human thought is inherently messy and error-prone, but it forces us to debate, to doubt, to test ideas – and to recognize the limits of our own knowledge. It builds self-reliance, both personally and collectively. For Kant, the use of reason was never solely about information; it was about enabling people to become agents of their own destinies, and to resist domination. It was about building a ethical society grounded in the shared principle of reason and debate, rather than unquestioning acceptance.

With all the undeniable benefits AI brings, the paramount challenge is this: how can we leverage its potential of superhuman intelligence without undermining human rationality, the bedrock of the Enlightenment and of free societies itself? That is likely one of the central dilemmas of our time. It is a question we would do well not to delegate to the algorithm.