Artificial intelligence could give more people access to knowledge than any previous technology. Yet it could also encourage a dangerous habit: accepting answers from an authority whose reasoning we cannot adequately examine. A society can become technologically sophisticated while losing its commitment to independent judgment.
That concern shaped a lecture by Harvard legal scholar Martha Minow at MIT on October 22. Speaking about democracy in the age of AI, she warned of a possible “loss of the Enlightenment.” Her concern involved the erosion of reason, evidence, tolerance, and participation.
Minow identified four foundations of democracy: education, reliable and accessible information, security from violence, and fairness. However, she also argued that technology could help strengthen democratic life. Her lecture, reported by MIT News on October 23, presented a warning about human choices rather than a prediction of inevitable collapse.
When an answer replaces an argument
One of the most valuable principles associated with the Enlightenment is that authority should face scrutiny. A claim deserves acceptance because evidence and reasoning support it. Prestige, tradition, and confidence cannot settle the question.
AI creates a particular temptation here. An answer can arrive immediately, expressed in polished language, with every hesitation removed. The reader may experience clarity without having examined whether the explanation deserves confidence.
Consider a student asking why a government collapsed. A useful response would distinguish established events from competing interpretations. It would identify uncertainty and provide evidence the student could inspect. A more dangerous response would deliver one seamless explanation that makes a disputed historical question appear settled.
The difference concerns how the reader learns to think. Does the exchange encourage investigation, or does it make investigation feel unnecessary?
A new authority for the skeptical mind
Secular people should recognize this danger. Rejecting religious authority does not automatically protect someone from intellectual dependence. A person can demand evidence from a priest and then accept an algorithm’s conclusion without applying the same standard.
The relevant question is whether we remain willing and able to challenge an answer. What supports it? What would count against it? Does the evidence establish the conclusion, or merely make it plausible?
These questions also apply when AI agrees with us. Agreement can make weak reasoning especially comfortable. A system that continually supplies persuasive defenses of our existing beliefs could turn apparent inquiry into an exercise in self-confirmation.
Freethought requires more than freedom from a particular religion. It requires the willingness to discover that our preferred answer is wrong.
Who controls the explanation?
There is also a political question behind the interface. If citizens increasingly rely on a small number of privately controlled systems to interpret public affairs, the organizations operating those systems could acquire considerable influence over public understanding.
Their choices would matter even without a deliberate propaganda campaign. Which questions receive detailed answers? Which sources receive attention? Which uncertainties remain visible? What can users discover about the rules shaping the response?
A democracy should demand ways to examine those decisions. Corporate assurances alone cannot establish that a system serves the public interest. Equally, handing governments unrestricted control over acceptable answers would create an obvious route to censorship.
The aim should be accountable power, independent scrutiny, and meaningful alternatives. Citizens need the practical ability to compare explanations and challenge decisions that affect them.
Education must preserve the work of thinking
AI could help a student understand a difficult passage, explore an objection, or practice constructing an argument. Those uses could widen access to education. However, completing an assignment and developing the ability it was designed to teach are different achievements.
If a student submits an argument they cannot explain, the finished document conceals the absence of learning. The same problem arises when a professional relies on a conclusion they cannot evaluate. Productivity becomes a misleading measure if nobody checks whether understanding has increased.
Education should therefore make room for explanation, revision, discussion, and the defense of conclusions. Students should learn to use powerful tools while retaining responsibility for what they claim.
That responsibility should grow with the consequences. A convenient answer about a film requires a different level of scrutiny from an assessment that could shape someone’s rights or livelihood.
Progress requires institutions, not faith
None of this requires treating AI as an enemy of reason. Tools that make evidence easier to inspect, reveal contradictions, and help people understand unfamiliar arguments could strengthen public debate. The opportunity is substantial.
But those benefits should become explicit goals. Systems should make uncertainty visible, support verification, and help users recognize the difference between a persuasive explanation and a well-supported conclusion. Public institutions should also preserve avenues for appeal when automated systems influence consequential decisions.
The Enlightenment was never a completed victory over prejudice or domination. Its principles demand continuing work, including scrutiny of technologies built through scientific progress itself. We should judge AI partly by whether it helps people perform that work.
A machine may produce an answer in seconds. A free society must preserve the right, the ability, and the habit of asking why anyone should believe it.
