top of page

The Art of Asking Better Questions in the Age of AI

Nov 9
4 min read

What a Cold War nuclear warning can teach us about critical thinking and discernment in a technological age


On September 26, 1983, Soviet Lieutenant Colonel Stanislav Petrov was on duty when a military computer announced that the United States had launched a nuclear missile toward the Soviet Union. The system then reported four additional launches. Petrov was given information that five American missiles were heading toward Soviet territory. Petrov had just minutes to evaluate the warning.


Petrov lived during the Cold War, a time that was dangerously divided. Major superpowers were jockeying for military power, political dominance, and nuclear supremacy. It seems that this particular computer was designed for this moment.


But Petrov was troubled by this moment. Not just because of the imminent threat, but because of the logic behind the computer's report. If the Americans intended to begin a nuclear war, why would they launch only five missiles? Wouldn't a genuine first strike involve a much larger attack? Wouldn't imminent threat have been signaled more clearly? These questions would prove to be life saving.


Although the satellite system was relatively new, and the warning had not been independently confirmed by ground-based radar. Petrov would had every reason to take the information seriously due to the temperature of the times, but he also had reasons to question it.


He weighed his questions more heavily than the computer's report. His decision was not to confirm an attack but to report a computer malfunction. Turned out he was right. Investigators later discovered that sunlight reflecting from high-altitude clouds had confused the satellite detection system.


We cannot know exactly what might have happened had Petrov made a different decision. Other military authorities would have been involved in any response so Petrov was only one person in a possible chain of decisions. Nevertheless, Petrov's judgment helped prevent a false warning from being treated as a confirmed attack.


And that raises an interesting question: What if Petrov had simply trusted the computer?


The Ancient Art of Asking Better Questions

Nearly twenty-five centuries earlier, the Greek philosopher Socrates had made something of a nuisance of himself by asking people to explain what they believed and why they believed it. His questions exposed assumptions, contradictions, and the sometimes uncomfortable distance between confidence and knowledge. And the youth in Greece correctly identified that his system of questions unnerved people, and also provided a means to get to the truth. Eventually Socrates was arrested and killed for impiety and corrupting the youth.


Of course, Socrates did not invent curiosity or questioning, of course, but the dialogues recorded by Plato demonstrate a distinctive approach to examining ideas. Rather than simply accepting a claim, Socrates pressed for definitions, investigated evidence, challenged assumptions, and considered the consequences of an argument.


Although Petrov was not conducting a Socratic dialogue, his reasoning illustrates several habits associated with Socratic questioning. He examined the assumptions behind the warning, evaluated the available evidence, and considered an alternative explanation.


These habits are useful far beyond military command centers. They are useful for both the adults and youth of today. We still need to teach students to examine arguments. Parents still need to understand how to reason with their children. Adults need to navigate problems and decisions in which the answers are not immediately obvious. And we all need to learn how to work with AI in a way that preserves critical thinking and not implicitly trust the assumptions people or computers make every day.


These skills are becoming even more valuable as technology makes information so readily available—and as artificial intelligence progresses from retrieving information to generating content, reasoning through complex problems, and increasingly acting on our behalf. This emerging capability, known as agentic AI, allows systems to pursue goals, make decisions about next steps, use tools, and carry out multistep tasks with varying degrees of human oversight. As technology moves from answering our questions to taking action, our ability to think critically, examine assumptions, evaluate evidence, ask better questions, and exercise sound judgment becomes not less important, but more essential.


When the Computer Sounds Convincing

Today, we have extraordinary access to information. Search engines and artificial intelligence can deliver answers in seconds. We can explore unfamiliar subjects, investigate historical events, and receive explanations of concepts that once required considerable time and effort to research.


These are remarkable tools, and I appreciate what they make possible. But I also recognize the temptation to accept a plausible answer without examining it. Students certainly do this, but adults are hardly immune. Sometimes the convenience of having an answer is simply more appealing than the work of evaluating one.


Artificial intelligence introduces an additional complication. Tools such as ChatGPT, Claude, Gemini, and even "personal assistants" designed for specific professional and/or personal tasks can produce remarkably clear explanations while occasionally getting the facts wrong. An incorrect date, a misunderstood question, or an unsupported conclusion may arrive in the same confident prose as an accurate answer. Computers, just like people, don't necessarily announce that something has gone wrong.


This is not an argument for abandoning technology. Petrov certainly didn't abandon his computer system. He considered its warning carefully, recognized reasons for doubt, and used his judgment. This is a warning to watch the sign of the times and respond with wisdom.


The challenge is learning when to trust an answer, when to investigate further, and when to ask a better question.


A human finger touches a glowing question mark on an AI decision interface, with branching paths leading to different outcomes.


Comments


bottom of page