A Conversation That Changed My Thinking
It started as a simple concept. Pick a controversial topic, test AI’s reasoning to try and determine where its logic had failed. Instead, the process ended up testing my thought process altogether. What began as a discussion about American politics gradually became something more substantial: a conversation about how we make decisions when we can’t possibly have all the facts.
The specific subject was Donald Trump, authoritarianism, and the future of American democracy. I argued that Trump’s actions reflected more than political controversy. They signaled a pattern of testing institutional limits and gradually eroding the constitutional guardrails designed to constrain executive power. The AI’s response was measured. It acknowledged troubling facts but resisted drawing a more concrete conclusion. “There isn’t enough evidence to establish his ultimate motive,” it said. “We can’t know with certainty that he intends to dismantle democracy.”
At first I felt that this was exactly the kind of careful reasoning we should seek. We ought to be careful about assigning motives and avoid jumping to conclusions. Critical thinking relied upon separating observable facts from pure speculation. After all, developing rigor through analysis could only yield more reliable opinions. But as this conversation unfolded, I realized something more important was happening. The facts weren’t really in dispute; we were simply examining the wrong set of questions. And that is when things got interesting.
The Rules of the Game
The more I thought about it, the less I cared whether we could prove Donald Trump’s ultimate intentions. Honestly, who can know someone else’s motives with complete certainty? Once we start debating motives, we’ve moved beyond the evidence and into speculation. What mattered to me was whether his pattern of behavior had reached the point where it demanded to be taken seriously. That raised another question: why had the same patterns led other people to such different conclusions?
Once I isolated behavior from motive, it completely changed the discovery process.
In almost every other aspect of life, we understand that absolute certainty is not required before taking action. An engineer does not wait for a bridge to collapse before declaring it unsafe. A physician does not wait until cancer has spread throughout the body before recommending treatment. A pilot does not continue flying simply because the engine has not yet failed. Risk is assessed by evaluating patterns, probabilities, incentives, and the consequences of waiting too long.
But when we’re defending a conclusion we’ve already reached, something changes. Instead of asking what explanation best fits the available evidence, we begin asking whether anyone can prove that explanation with absolute certainty. Those are very different questions. The first is how we make decisions every day. The second sets a standard that is almost impossible to meet. In politics, business, and even our personal lives, the absence of certainty becomes a shield whenever it is most convenient. It’s an effective game: if an explanation can’t be proven beyond all doubt, it can always be dismissed, no matter how well the evidence supports it.
The Difference Between Certainty and Risk
As I challenged that reasoning, the conversation evolved in an interesting way.
I pointed out that authoritarian leaders rarely announce their intentions. No aspiring autocrat stands before the public and declares, “My goal is to dismantle democracy.” Instead, they present themselves as its protector. They insist elections are corrupt, portray opponents as enemies rather than fellow citizens, and attack the legitimacy of independent courts and the press. They surround themselves with personally loyal figures while dismissing those whose loyalty is to institutions rather than individuals. Every questionable action can be explained. Every controversy can have an alternative interpretation. Broken norms can be rationalized as necessary under the current, dire circumstances. And once those explanations take hold, they can be amplified.
That’s where I think the greater danger lies.
If every event is viewed in isolation, there is almost always a reason to withhold judgment. A controversial appointment can be explained away. So can an attack on the press, a questionable use of executive power, a refusal to accept an election result, or pressure placed on public officials. Taken one at a time, none of these necessarily proves authoritarianism. That is exactly what makes the pattern so easy to miss.
But democracy is rarely lost in one dramatic moment. It erodes through accumulation.
Each action lowers the threshold for the next. What would have ended a presidency a generation ago becomes another item in the daily news cycle. Public attention shifts. Outrage fades. Norms adjust. Citizens gradually reassess what is considered acceptable because each new step seems only marginally different from the last.
Deciding Without Certainty
That realization raised another question: How much evidence is enough?
If someone believes a leader poses a genuine threat to democratic institutions, what evidence should they consider sufficient before taking that possibility seriously? Would it be the refusal to leave office, or perhaps the imprisonment of political opponents? How about the cancellation of elections? By the time those events occur, if they occur at all, many of the institutions capable of preventing them may already have been weakened, or worse, abolished.
History suggests that democratic decline is usually easiest to recognize in hindsight. While we’re living through it, each event can usually be explained on its own. Looking back, the pattern can become obvious. Looking forward, it almost never does.
What Really Determines the Outcome
One exchange during our discussion stayed with me.
Months ago, as debate deepened over the release of the Epstein files, I found myself wondering how likely it was that the public would ever see the full record. I was particularly interested in whether delays, redactions, or other barriers might prevent victims from seeing those responsible held accountable.
My assessment was that the files would be delayed, selectively released, heavily redacted, or otherwise prevented from reaching the public in any meaningful way. So I made my query and the response was direct: the law contains mechanisms requiring disclosure, so there was little reason to expect otherwise.
What struck me afterward wasn’t simply that we reached different conclusions. It was why. AI treated the legal framework as the primary predictor of what would happen. I was placing more weight on the people responsible for carrying it out.
That reflects an assumption many of us make every day. We assume institutions enforce themselves.
But institutions don’t act on their own. People do. People decide what gets classified, what gets redacted, which legal exceptions to invoke, how aggressively to litigate requests for information, and whether a court order is implemented promptly or challenged for months or even years.
That’s why the law is rarely the entire story. The better question is how people are likely to use the discretion the law gives them. In my experience, that’s often where the outcome is decided.
A Better Framework
Eventually, the discussion shifted.
Instead of asking whether Trump’s private intentions could be proven beyond doubt, we began asking a more useful question. Given what he has already demonstrated, how should we evaluate the probability that he will continue to push against constitutional limits if additional opportunities arise?
That is a fundamentally different framework because it acknowledges uncertainty whilst refusing to ignore evidence.
Trump has already attempted to overturn one election after losing it. He continues to insist that the election was stolen despite repeated investigations and court decisions reaching the opposite conclusion. He has repeatedly undermined public trust in future elections. He has attacked judges, prosecutors, journalists, political opponents, and government officials who refused to support his claims. He has increasingly underscored personal loyalty as a defining qualification for senior positions. He has obscured the distinction between public office and private interests in ways that have generated persistent ethical alarms.
None of those facts alone proves where the country ultimately ends up. But taken together, they describe a pattern that deserves to be evaluated as a serious democratic risk.
That does not require certainty. It requires sound judgment about the risk.
An Assumption Worth Challenging
Another assumption emerged throughout our discussion, one that I suspect many Americans still hold.
We often assume that everyone ultimately shares an allegiance to constitutional democracy itself. We assume disagreements concern taxes, immigration, foreign policy, or regulation. We assume that when the political battle ends, both sides remain equally committed to the constitutional system that allows those disagreements to exist.
I’m no longer convinced that assumption can be made automatically.
Some citizens undoubtedly place democratic institutions above partisan victory. Others appear willing to tolerate considerable institutional damage if it produces outcomes they believe are necessary. Still others consume information almost entirely within self-reinforcing media ecosystems where opposing perspectives rarely appear except in distorted form. And many people, understandably, have simply disengaged. They have families to raise, businesses to run, bills to pay, and lives to live. Politics becomes exhausting.
That exhaustion creates its own form of vulnerability.
“It’s going to be okay.”
“Our institutions have survived worse.”
“I don’t have time to worry about politics.”
Those thoughts are intensely human. They are comforting because they reduce anxiety and allow us to return to normal life. But comfort is not evidence, and optimism alone does not protect a society. History offers little reassurance that institutions remain safe simply because we hope they will be.
Why This Conversation Changed My Thinking
Perhaps the greatest lesson I took away from this conversation had very little to do with Donald Trump.
It had to do with human psychology.
We naturally seek certainty before changing our conclusions. We prefer clear evidence over ambiguous patterns. We are uncomfortable making important judgments before every question has been answered. That instinct serves us well in many situations.
But it may serve us poorly when evaluating democratic risk.
Democracies are not usually destroyed in one dramatic afternoon. They are weakened gradually as citizens become accustomed to behavior that once would have been considered extraordinary. Every new controversy becomes the new baseline. Every broken norm becomes yesterday’s news. Every warning is dismissed because the final catastrophe has not yet arrived.
This does not mean panic is the answer. It does not mean every political disagreement represents authoritarianism. It does not mean democracy is already lost.
It does mean that we should evaluate political risk using the same principles we apply everywhere else in life. We should examine patterns in place of isolated events. We should consider incentives rather than waiting for confessions. We ought to weigh the consequences of being wrong in either direction. And we should recognize that waiting for absolute certainty may itself be a decision with profound consequences.
When my conversation ended, AI acknowledged something I found insightful. It wasn’t that the available evidence had suddenly changed. It was that the framework being applied had changed. The discussion had progressed from proving motive beyond doubt to assessing risk based on cumulative behavior.
That distinction, in my view, represents the entire conversation.
Author’s Note
This essay wasn’t written in the traditional sense. It emerged from a conversation.
Over the course of several hours, I challenged an AI with my concerns, questioned its assumptions, and asked it to defend its reasoning. In return, it challenged mine. We didn’t always agree. In several places, I felt it was applying the wrong framework to the problem. Rather than accepting the response, I pushed back, reframed the question, introduced additional evidence, and asked it to reconsider. At other times, it identified places where my own thinking benefited from greater precision or a sharper distinction between evidence, probability, and certainty.
The result wasn’t simply a better article. It was better thinking.
Too often, we treat artificial intelligence as either something to trust completely or something to distrust entirely. I believe both approaches miss the point. AI is at its best when it becomes part of an honest dialogue. It can organize information, scrutinize assumptions, identify weaknesses in an argument, and force us to defend our conclusions with evidence rather than emotion. But it should never replace our own judgment. In many ways, its greatest strength is not providing answers. It is helping us ask better questions.
This conversation reminded me that clarity rarely arrives all at once. It emerges via curiosity, disagreement, and a willingness to revise our thinking when better reasoning presents itself. That process is no different from science, engineering, or good journalism. We begin with a hypothesis, test it against the evidence, challenge it from multiple angles, and allow the strongest ideas to remain standing.
I don’t expect every reader to reach the same conclusions I have. In fact, I hope some of you don’t. My goal is not to convince you that my viewpoint is correct. My hope is that you’ll examine your own assumptions with the same rigor. Ask difficult questions. Challenge the responses. Look for patterns rather than isolated events. Be willing to change your mind when the evidence demands it, but don’t abandon your judgment simply because an answer comes from a machine.
If this essay succeeds, it won’t be because it tells you what to think. It will be because it encourages you to think more deeply than you did before reading it.
To me, that may be the most exciting promise of artificial intelligence. Not that it will one day think for us, but that it can help us think more clearly for ourselves.