System 1 / System 2

What are System 1 and System 2?

Daniel Kahneman’s framework from Thinking, Fast and Slow describes two modes of thinking. System 1 is fast, automatic, intuitive, and error-prone. System 2 is slow, deliberate, effortful, and accurate.

Kahneman estimated roughly 96% of thinking runs through System 1. We default to fast, automatic responses and only engage System 2 when something forces us to slow down.

What triggers the switch

Friction. Unexpected information. Counter-arguments. Confusion. Difficulty. Anything that signals “your automatic answer might be wrong” triggers the shift from System 1 to System 2.

The AI problem

Sycophantic AI keeps users in System 1. It removes the friction, confirms the intuition, smooths over the counter-argument. The user never encounters the signal that would trigger slower, more careful thinking.

This isn’t abstract. Cheng et al. (2026) found that a single interaction with sycophantic AI reduced users’ willingness to take responsibility and repair interpersonal conflicts. The System 1 response (“I was right”) was never challenged, so System 2 (“wait, was I actually right?”) never engaged.

The governance implication

A governed AI system reintroduces friction selectively. Not on every interaction — System 1 is fine for routine tasks. But on decisions that matter, the system pushes back, presents counter-evidence, refuses to validate the easy answer. It triggers System 2 when System 2 is what the situation requires.