Skip to main content

New Research, Same Old Human Problem: We’d Rather Sound Smart Than Actually Think

New Research, Same Old Human Problem: We'd Rather Sound Smart Than Actually Think graphic on cognitive surrender and critical thinking at work

New Research, Same Old Human Problem: We’d Rather Sound Smart Than Actually Think

There’s a new paper out of Wharton—Thinking, Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender —and it politely confirms something I (and you) have been watching play out in real time:

We are getting real comfortable letting AI do our thinking.

They’ve got a clean name for it now:

“Cognitive surrender—adopting AI outputs with minimal scrutiny.” (p. 2)

Now I’m going to translate that into plain English from an older woman who has sat in more boardrooms and conference rooms than she cares to count:

You didn’t think it through. You just liked how it sounded. Good ‘enuf..

We Now Have Three Brains in the Room—and One of Them Doesn’t Work for You

The research lays it out like this:

System 1: your gut System 2: your reasoning System 3: AI

And here’s the part people should not skim past:

“System 3… can supplement or supplant internal processes.” (p. 2)

Supplant. That’s a real polite academic way of saying: It will take over if you let it.

And right now? A whole lot of folks are letting it.

Let Me Show You How This Actually Plays Out

Change Consulting (Where “Sounds Right” Will Get You in Trouble)

You’re working a utility transformation—big change, entrenched culture, everyone smiling in meetings and resisting in the hallway. You run AI to draft your change strategy, and it gives you a beautiful plan: stakeholder alignment, comms cadence, phased rollout. Clean. Logical. Boardroom ready. And you send it. What you didn’t do was pressure-test it against how decisions actually move in that organization. Because here’s the truth: AI gave you a good answer. It just didn’t give you their answer. That gap? That’s where your credibility goes to die.

Electrical Construction (Where “Close Enough” Can Cost You Real Money)

Project’s tight. Substation upgrade. Schedule’s already breathing heavy. PM runs AI for risk mitigation—gets a tidy list: sequencing, supplier coordination, safety overlays. Looks solid, so it gets adopted. Nobody slows down long enough to notice it ignored a site constraint—access limitations that change everything about crew timing. Plan moves forward anyway. And now you’ve got a problem that didn’t come from bad intent. It came from borrowed thinking.

Leadership Decision (Where We Pretend a Summary Is Judgment)

You’re helping make a senior hire. You drop interview notes into AI and it hands back a clean summary: strong strategist, moderate execution risk, high alignment. It’s neat. It’s confident. It feels decisive. And you go with it. What you didn’t do was go back and wrestle with the messy parts—the contradictions, the tone, the hesitation. You accepted a summary as a decision. That’s not efficiency. That’s outsourcing judgment.

Now Here’s the Part That Should Make You Sit Up Straighter

The research didn’t just describe this—they measured it:

“Accuracy significantly rose when AI was accurate and fell when it erred.” (p. 2)

Meaning: You don’t just get smarter with AI.

You get as smart—or as dumb—as it is in that moment.

And then this little gem:

“Engaging System 3 also increased confidence, even following errors.” (p. 2)

So now let’s paint the full picture:

You’re wrong. You sound right. And you’re confident.

I have seen entire executive teams run off that combination.

Why This Happens (And No, It’s Not Because You’re Lazy)

AI is:

Fast Polished Certain-sounding

And we are wired to trust things that sound like they know what they’re doing.

So instead of triggering more thinking, it does the opposite.

“The decision-maker no longer constructs an answer, but adopts one.” (p. 4)

That’s the shift right there.

From thinking → to accepting.

There’s a Line Here—and Most People Don’t Know They Crossed It

The research makes a clean distinction:

Cognitive offloading = using AI to support your thinking Cognitive surrender = letting AI replace your thinking

“A relinquishing of cognitive control… substituting it for their own reasoning.” (p. 4)

One makes you better.

The other just makes you faster at being wrong.

The Part That Worries Me (And Should Worry You Too)

This isn’t loud. It doesn’t show up as obvious failure.

It shows up as:

Smart-sounding recommendations that don’t land Plans that look right but don’t work in the field Decisions that feel clean because nobody actually wrestled with them long and hard enough to find the holes. Wording that doesn’t quiet (or is it quite) land with these folks

And the most concerning line in the whole paper:

“Users may not know when or why they have deferred.” (p. 17)

That’s it. That’s the whole problem. You think you’re thinking.

You’re not.

So, What Do You Do (Because We’re Not Going Back to Whiteboards and Guesswork)

You don’t ditch AI.

That would be dumb.

But you do this:

Pause when you instantly agree If it feels easy, that’s your signal to slow down. Run one clean pass without it Just to see if you land in the same place. Ask what’s missing, not what’s good AI is great at sounding complete. That doesn’t mean it is. Own the thinking If your name is on it, your brain better have been involved.

And for the record—the research shows when people are given incentives and feedback, they override bad AI more often (p. 34–35) . Meaning this isn’t inevitable. It’s behavioral.

Final Thought (From Someone Who’s Seen a Few Cycles of “The Next Big Thing”)

AI is powerful.

It will make you faster. It will make you sound sharper. It will absolutely make your work cleaner.

But it will also, very quietly, try to take over the part of your job that actually matters: Judgment.

And if you’re not paying attention, you’ll slide from: decision-maker → reviewer → approver

…without ever realizing you handed over the wheel.

That’s cognitive surrender.

And I don’t care how good the tool is—That’s a trade you don’t want to make.

Quick Answers

What is cognitive surrender?

A term from Wharton research for adopting AI outputs with minimal scrutiny. Instead of constructing an answer, the decision-maker adopts one, often without knowing when or why they deferred.

What is the difference between cognitive offloading and cognitive surrender?

Cognitive offloading uses AI to support your thinking. Cognitive surrender lets AI replace it. One makes you better, and the other just makes you faster at being wrong.

How can leaders avoid cognitive surrender with AI?

Pause when you instantly agree with an AI answer, since easy agreement is a signal to slow down, and run one pass without the tool to compare. The research found people override bad AI more often when they have incentives and feedback.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646

https://dx.doi.org/10.2139/ssrn.6097646

Originally published on LinkedIn on May 1, 2026. Read it on LinkedIn.

Dana Houston Jackson

Dana Houston Jackson is the founder of Change With A Coach Approach, LLC and the author of The AI Adoption Traps. She started in manufacturing, spent 10 years in construction, taught K-12 and university students, and ran technology projects as a scrum master and project manager. As a consultant she has stood up project management and change management offices and trained clients across industries, including work with the Department of Energy. She’s a speaker, advisor and executive coach with 30 years in the work, now focused on helping leaders turn AI activity into lasting capability.

    Comments are closed