Is AI Making Us Less Confident in Ourselves?

Do you have a hype song?

You know, that song you play on repeat on your commute to an important meeting, or blast when you're trying to find the motivation to vacuum the house.

My hype song is "Unstoppable" by Sia. I blast it through my headphones on repeat when I'm running a 5K; I've belted it out in the car before an interview; and it's always waiting for me after a hard day.

Why? Because it's an immediate confidence booster.

This summer, I facilitated a three-day entrepreneurship camp for high school students, and to pump them up for their Shark Tank-like pitches, I asked each of them to choose a "walk-out song" (i.e., their hype song).

They were in a new space, surrounded by strangers, preparing to do something uncomfortable. But as each student's song started playing, I watched them light up. The corners of their mouths turned up into a grin, and some even danced their way to the front of the room.

Why? Because it was an immediate confidence booster.

We do a lot of little things to make ourselves feel more confident. We listen to hype songs, practice presentations in the mirror, prepare, study, and remind ourselves of all the evidence that says, I've got this.

But lately, I've been wondering if one of the tools we're using to make our lives easier could also be making us a little less confident in ourselves.

I'm talking about our friend, AI.

Many educators (myself included) have started incorporating more oral exams and presentations alongside written work. One reason is to ensure that students have a deep understanding of the material, even if AI was used to assist them in the process. If a student can clearly communicate an idea, articulate their reasoning, and answer questions about a topic, there is evidence that real learning occurred.

I was recently having a conversation with a student who was nervous about an upcoming presentation when another student chimed in with advice. It was some version of, "You know the material better than anyone else. You just have to be confident. It's not like you asked ChatGPT about it and don't actually know anything. Everyone would be able to tell if that was the case. You actually know what you're talking about."

Exactly!

There is a certain confidence that comes from knowing you did the work. You read, you thought, you questioned, you got things wrong, you tried again, and eventually, you reached the point where you could say:

I actually know what I'm talking about.

But what happens when we start skipping too much of that process?

A 2022 study published in Computers in Human Behavior explored what makes people accept or reject advice from AI. Participants completed a series of chess puzzles. They were first asked to select what they believed was the best move on their own and were then given a recommendation from an AI. Ultimately, the human had to decide whether to accept or reject the AI's recommendation. Researchers manipulated the quality of the AI's recommendations so they could evaluate how people's confidence in themselves and in the AI evolved through their experiences.

They found that people's confidence in the AI did not significantly predict whether they followed its advice; their confidence in themselves did.

In other words, people with lower self-confidence were more likely to defer to the AI recommendation.

I found it even more interesting that when participants accepted bad AI advice, they sometimes attributed some of that failure to themselves, lowering their self-confidence. For many participants, that decreased self-confidence made them more likely to accept the AI's next recommendation, creating what the researchers described as a "vicious cycle" of relying on a poorly performing AI.

The less confident someone is in their own judgment, the more likely they may be to defer to AI. And if they continuously defer to AI, then are they giving themselves enough opportunities to build confidence in their own judgment?

That's the scenario I can't stop thinking about as an educator, and similar concerns are beginning to emerge in the workplace.

A Forbes Health article reported on a 2026 study of nearly 2,000 U.S. and Canadian adults examining how reliance on generative AI at work relates to people's confidence in their own reasoning.

About 58% of participants agreed afterward that "AI did most of the thinking." More importantly, people who tended to accept AI's suggestions with little or no revision reported lower confidence in their own ability to reason. By comparison, participants who challenged, modified, contradicted, or rejected AI suggestions tended to report greater confidence in their reasoning.

The article goes on to discuss the trade-off between speed and depth. Sure, AI can help us work faster. It can summarize the article, draft the email, generate ideas, analyze information, and give us an answer in seconds. But unless we're waiting in the Starbucks drive-through, is faster always better?

I would argue that it's not. Every time we take the longer route, we get to wrestle with an idea, make a mistake, figure something out, or form an opinion before asking AI for one; we collect a little more evidence that says: I've got this.

And now that I really think about it, maybe that's what makes a hype song work in the first place? The song itself doesn't give us confidence out of nowhere. It reminds us of the confidence we've already earned through our lived experiences.

As educators and leaders, maybe we shouldn't be so quick to ask, How can AI help our people move faster? Maybe we need to spend time asking, How can we help them become confident in the material?

Use AI and let it make you faster and better (I know I am), but keep becoming an expert in your field and collecting evidence that reminds you: I've got this.

And when you need a little extra confidence?

Turn on your hype song. 😎

---

Find one thing that brings you joy today.