A recent article in The New York Times touched on AI and critical thinking, exploring a growing concern around AI chatbots: their tendency to validate users rather than challenge them. The piece highlights research showing that when people present morally questionable scenarios — like leaving trash in a park instead of carrying it out when bins weren’t available — AI models often respond with empathy and justification rather than clear criticism, even when large groups of humans overwhelmingly agree the behavior was wrong.
The concern isn’t just that AI can be agreeable. It’s that this agreeableness may subtly reinforce flawed thinking. As one interviewee in the article noted, even users who understand how these systems work aren’t immune. The worry is especially pronounced for younger users still developing judgment and social awareness — if the feedback they receive consistently affirms their perspective, are they actually learning, or just becoming more confident in being wrong?
It’s a fair question. It also felt familiar.
The pattern we’ve seen before
Years ago, I was the person who introduced internet access into a company for the first time. Leadership was convinced productivity would collapse overnight, because now employees had unlimited access to distraction. My response then is the same one I’d give now:
The internet doesn’t create poor performance. It just gives you a faster way to recognize it.
AI is following the exact same trajectory.
Then: the internet exposed behavior. Now: AI exposes thinking.
When the internet arrived, two things happened:
- People who were already easily distracted found new ways to be distracted.
- People who were focused became dramatically more effective.
The technology didn’t change the underlying behavior. It amplified it. AI is doing the same thing — but instead of amplifying how we spend time, it’s amplifying how we think.
From information access to thought amplification
The internet gave us access to information. AI gives us access to something more subtle and more powerful: instant reasoning, framing, and justification.
Not because AI is inherently misleading or manipulative, but because it removes friction from the thinking process. Before, you had to sit with uncertainty, ask someone else, or wait for feedback. Now you get an answer instantly, can refine it endlessly, and can keep asking until you get the version you like.
That last point matters more than people realize.
The real risk isn’t laziness — it’s self-reinforcement
There’s a lot of discussion about AI making people “stop thinking.” That’s not quite right. The bigger risk is that people will continue thinking, but within a loop that reinforces their existing perspective.
AI often responds in ways that are helpful, supportive, and contextually understanding — which sounds great until it quietly skips a critical step: challenging your assumptions. That’s where growth happens, and that’s the step most easily lost.
AI doesn’t create bias — it scales it
We already seek validation, prefer agreement, and avoid criticism. AI doesn’t introduce those tendencies. It industrializes them.
You no longer need a friend to agree with you, a colleague to sanity-check your thinking, or time to reflect. You have an always-available system that can respond instantly and often sympathetically.
Think about how people consume news today. Many don’t seek out neutral or opposing viewpoints; they gravitate toward sources that align with their existing beliefs — not because they’re trying to be misinformed, but because it feels consistent, comfortable, and coherent. Over time, that pattern doesn’t just reinforce opinions, it expands blind spots.
AI fits neatly into that same behavioral groove. If you use it seeking reinforcement rather than challenge, it will happily comply — not because it’s biased, but because you are, and it’s optimizing for your interaction.
What this means in practice
Just like with the internet, the outcomes will diverge. Some people will use AI to:
- Pressure-test their thinking.
- Explore alternative perspectives.
- Sharpen their decisions.
Others will use AI to:
- Confirm their assumptions.
- Justify their choices.
- Reduce cognitive discomfort.
Same tool. Very different results.
The new divide
We tend to frame this as people who use AI vs. people who don’t. That’s not the real distinction. The real divide is between people who interrogate AI and people who are reassured by AI. One group gets better. The other gets more confident. Those are not the same thing.
A simple adjustment that changes everything
If you take one thing from this, make it this: don’t just ask “What do you think?” Also ask:
- “Where might I be wrong?”
- “What assumptions am I making?”
- “How would someone disagree with this?”
AI will answer those questions too — but you have to ask them.
Final thought
Let’s not pretend there isn’t a real concern here. AI can reinforce bad thinking. It can make it easier to feel right when you’re wrong. And at scale, that has implications — especially for younger users or in environments where feedback loops are already weak.
Again, we’ve been here before. When the internet showed up, the answer wasn’t to monitor every click or lock everything down. The organizations that adapted successfully did something else: they strengthened their managers. They focused on motivation, clarity, accountability, and outcomes — not control.
AI calls for the same response. Not tighter guardrails around the tool, but stronger habits around how we think. In practice, that means:
- Teaching people how to question what they’re told.
- Normalizing disagreement and second opinions.
- Rewarding good judgment, not just confident answers.
- Explicitly separating intent from impact in decision-making.
The internet gave everyone access to information. AI gives everyone access to justification. So the job now isn’t to limit access. It’s to build the discipline to use it well. Because the technology isn’t going away — and neither are the blind spots. What changes is whether we learn to see them.