Motivation & Habit

What actually moves your behaviour, and how the brain automates a repeated action so it stops costing you willpower to start. The subtraction at the centre of it is the same one your reward model learned in Fine-tuning & Alignment.

Before you start. The last lesson, Attention & Cognitive Load, showed that your working memory is a small tank that fills up fast under focused thought. This one is about the trick the brain uses to run a behaviour without touching that tank at all: it automates the behaviour into something called a habit, and this lesson is about how.

The idea in 30 seconds

In 1954 a rat given a lever that sent a small electric pulse into one spot of its brain pressed it thousands of times an hour, and would starve rather than stop.
Dopamine does not track how good something is. It tracks the gap between what happened and what you expected, a subtraction called reward prediction error.
A reward you can fully predict stops producing much signal at all, which is why a first win feels bigger than the hundredth identical one.
A cue triggers a craving, the craving drives a routine, and the routine’s reward teaches the brain to run the whole thing again: that cycle is a habit loop.
You cannot switch off a craving directly, but you can change the cue that starts it and the friction on the routine it drives.
A reward on an unpredictable schedule never lets that gap close, which is why it is harder to put down than one that pays off every single time.
A cue gets its pull by classical conditioning: a neutral signal repeatedly paired with a reward starts to trigger the craving on its own, and ignoring it later suppresses that link rather than erasing it.

Read the story · Skip to the numbers