Acting on autopilot

A note for Lesson no. 4, "Your Brain Predicts (Almost) Everything You Do," in Seven and a Half Lessons About the Brain by Lisa Feldman Barrett.
Some context from page 77 is:

Think about the last time you acted on autopilot. Maybe you bit your nails.

What is this?

Acting on auto-pilot is called a "habit." Cues in the world, or perhaps in the body, lead you to perform an action without considering the costs or consequences. Traditionally, habits have been understood as behaviors that are triggered by an object in the world without consideration for the outcome (i.e., a stimulus-response association, like Pavlovian or classical conditioning), and recently there has been a return to this sort of view.

At the same time, some scientists have hypothesized that habits arise from something called "model-free" reinforcement learning. Model-free, in this terminology, does not mean "prediction-free." It is a set of mathematical algorithms that predict which action a creature will take based on the expected immediate return on the investment (i.e., the immediate value of the action, or action-outcome associations).

These two competing views can be nicely integrated into a predictive processing view of the brain, which is, in turn, consistent with other research on habits. For example, habits arise in situations where it is easy to predict, based on the current contextual cues (i.e., the "stimuli"), a consistent, expected cost of an upcoming action (i.e., the "response") as well as a consistent, expected reward (i.e., the "outcome"; e.g., see ). That is, habits may arise in situations where little prediction error is expected and therefore is not detected when present, leading to a maintenance of the habit.