Does Your Brain Have an 'Autopilot Mode'? Science Now Knows How to Measure It

Ever driven home without remembering the route? Or frozen facing a new problem, like assembling furniture without instructions? Behind these moments lies Rasmussen's famous S-R-K model: **Skill** (automatic ability), **Rule** (known rules), and **Knowledge** (deep reasoning for the unknown). Until now, it was just theory. Now, thanks to EEG caps and artificial intelligence, researchers have "photographed" it in the brain.

They asked volunteers to perform three tasks: tracking a moving target (Skill), navigating with fixed rules (Rule), solving never-seen-before situations (Knowledge). Meanwhile, the electroencephalogram recorded **Global Field Power (GFP)**, the overall energy of brain waves, band by band. Then, a machine learning algorithm cross-referenced this data with **Brodmann areas**, the historic maps of the cortex.

The result? The model recognizes in real-time which "mode" the brain is operating in. The motor and sensory cortex lights up for Skill; prefrontal and parietal areas engage for Rule and Knowledge, with increasing cognitive load. It's the first time theory becomes objective measurement.

Why does it matter? Imagine safety systems in nuclear plants or cockpits that "sense" if the operator is on autopilot or improvising, preventing errors. Or brain-computer interfaces adapting to your fatigue level. The mind is no longer a black box: we are learning to read its source code.

The study, published in *Brain Sciences*, addresses a historical gap in human factors engineering: the lack of quantitative metrics for Rasmussen's Skill-Rule-Knowledge (S-R-K) model. The methodology relies on high-density EEG acquisition during three experimental paradigms designed to isolate the three levels of cognitive control: a visuomotor tracking task for Skill (automatic, bottom-up control), a rule-based navigation for Rule (schema-based, if-then control), and problem-solving in an unknown environment for Knowledge (analytical, top-down control).

Analysis of **Global Field Power (GFP)** — the standard deviation of the electric potential across all electrodes, a proxy for globally synchronized neural activation — revealed distinct spectral signatures across frequency bands (theta, alpha, beta, gamma) for the three conditions. Mapping onto **Brodmann Areas (BA)** showed: primary activation of BA 4 (motor cortex), BA 6 (premotor), and BA 1/2/3 (somatosensory) for Skill; involvement of BA 7 (superior parietal), BA 40 (supramarginal gyrus), and BA 9/46 (dorsolateral prefrontal) for Rule; massive recruitment of BA 10 (anterior prefrontal), BA 9/46, and bilateral fronto-parietal networks for Knowledge, consistent with literature on *cognitive control* and *working memory*.

The Machine Learning classifier (likely SVM or Random Forest with leave-one-subject-out cross-validation, given the context), trained on GFP spectral features per BA, achieved significant accuracy in discriminating the three states. This paves the way for **adaptive neuroergonomics**: systems monitoring cognitive load in real-time to prevent *skill-based slips* (lapses in automatic mode) or *knowledge-based mistakes* (planning errors), critical in aviation, operating rooms, and autonomous driving.