A new academic paper from Wharton was recently published on how AI is changing the way we think and I’m convinced it names something hugely important for anyone working in employee experience, internal communication, leadership or organisation design.
Not because it tells us AI is dangerous.
Not because it predicts some dystopian future.
But because it asks a more subtle question:
What happens to human thinking when AI becomes part of everyday work?
The paper is called Thinking Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender. And once you’ve read it, it’s hard not to see your organisation, and your own ways of working, a little differently.
The idea, in plain English
Most of us are familiar with the idea that humans think in two main ways (popularised by Daniel Kahneman in his seminal text Thinking, Fast and Slow)
- Fast, intuitive thinking – instinctive, emotional, pattern-based
- Slow, deliberate thinking – reflective, analytical, effortful
The authors of the research argue that this model is now incomplete.
When we work with AI, we introduce a third system of thinking, one that doesn’t live in our heads at all. They call this System 3: artificial cognition.
AI doesn’t just support human thinking. In many situations, it actively participates in it and sometimes replaces it.
This matters, because it changes how decisions are made, not just how fast they’re made.
What is “cognitive surrender”?
One of the paper’s most important contributions is via the concept of cognitive surrender.
Cognitive surrender in this context happens when:
- An AI system produces an answer
- The answer sounds confident, fluent and authoritative
- And we accept it with little or no scrutiny
Crucially, this is not the same as using tools well.
The authors make a clear distinction between two very different behaviours:
Cognitive offloading (helpful)
Using AI to:
- Reduce effort
- Handle complexity
- Generate options
- Support human judgment
Here, the human is still thinking. AI is extending cognition, not replacing it.
Cognitive surrender (more risky)
Deferring judgment to AI and adopting its output as your own, with minimal evaluation.
In this state, human thinking doesn’t disappear, it just quietly steps aside.
Across multiple experiments, the researchers found that people frequently followed AI advice, even when the AI was wrong. More strikingly, people often became more confident after consulting AI, regardless of whether the answer was correct.
Accuracy started to track AI accuracy, not human reasoning.
This is really about how work is designed
One of the most important things about this research is what it doesn’t say.
It doesn’t blame individuals.
It doesn’t frame this as a lack of intelligence.
It doesn’t suggest people are naïve or careless.
Instead, cognitive surrender is shown to be a predictable response to the conditions many organisations create, especially now AI is embedded into everyday work:
- Constant time pressure
- High cognitive load
- Increasing complexity
- A push for speed and certainty
- Systems that reward outputs, not reasoning
- Tools that present answers as clean and final
In those conditions, not thinking deeply can feel like the most rational option.
This is a design issue, not a motivation issue.
Can incentives and feedback help?
The researchers also explored whether cognitive surrender could be reduced.
When accuracy mattered, feedback was immediate, and consequences were visible, people were more likely to challenge faulty AI outputs. Incentives and feedback helped re-engage human judgment.
But, and this is key, cognitive surrender didn’t disappear. It reduced, but it didn’t go away.
That tells us something important:
You can’t fix this with nudges alone.
You have to look at how work, systems and expectations are designed.
Why expertise still matters (more than ever)
The paper shows that people with higher cognitive capability and a stronger tendency to engage in reflective thinking are significantly better at spotting and overriding faulty AI. This has important implications for expertise: without the underlying capability to evaluate, challenge and sense-check outputs, it’s hard to know when not to trust the system.
This directly challenges the lazy narrative that:
“AI makes expertise irrelevant.”
In reality:
- AI raises the floor
- Human expertise raises the ceiling
- And expertise is what prevents blind trust
Without domain knowledge, confidence and judgment, people don’t know when to challenge AI or even that they should.
This is where the real risk lies: not in AI replacing experts, but in organisations quietly deskilling their people by design.
Why this matters
This research matters because it shifts the conversation away from AI capability and towards how organisations shape human thinking.
When AI becomes fluent, fast and confident, the limiting factor is no longer access to information. It’s judgment. And judgment doesn’t live in individuals alone. It is shaped constantly by:
- how work is designed
- how decisions are made and explained
- how much time and space people have to reflect
- whether challenge is encouraged or quietly discouraged
- whether consequences are visible or abstract
In environments optimised for speed, certainty and output, reflective thinking is often the first thing to disappear. Under those conditions, deferring to AI can feel not just efficient, but sensible.
This is why employee experience and internal communication matter so much in an AI-augmented organisation, not as “soft” disciplines, but as enablers of judgment.
Experience design determines whether people are given the conditions to think: the pace of work, the visibility of consequences, the clarity of decision rights, and the permission to pause and question.
Communication shapes how people make sense of what they see: how uncertainty is framed, whether reasoning is made visible, and whether human judgment is still expected, even when systems produce confident answers.
Seen through this lens, communication is not just about sharing information. It becomes part of the organisation’s cognitive infrastructure: the narratives, signals and norms that influence when people trust, when they challenge, and when they stop thinking altogether.
People-first approaches have always argued that employees are not passive recipients, but active meaning-makers. This research extends that idea into the AI era. When artificial systems can generate answers instantly, sense-making becomes more important, not less.
The real risk is not that AI replaces human thinking, but that organisations unintentionally design work in ways that no longer ask people to think at all.
Why this feels so important right now
This paper isn’t anti-AI. In fact, it shows clearly that AI can be incredibly useful, especially under pressure. But it also surfaces a quiet shift that should concern anyone designing organisations for the long term: If we don’t intentionally design for cognitive offloading, we will default into cognitive surrender.
Not because people don’t care.
But because the system no longer asks them to think.
For me, that makes this one of the most important pieces of research I’ve read recently – not because it predicts the future, but because it helps us design for it more wisely.
If you’re interested in how to design work that keeps human judgment alive while making the most of AI, drop us a line this is exactly the kind of challenge we love working on.
The full paper is well worth reading if you want to explore the research in more depth – you can find it here
@Emma
