The book lay open on the kitchen table, spine cracked, a testament to the hours she'd spent devouring its pages. Her own life, laid bare for the world to see. A stranger’s name was on the cover, a woman she’d never met, but the words… the words were *her*. She ran a hand through her hair, tugging at the strands, a nervous habit that had returned with a vengeance.

She’d left the house early that morning, pretending to go to the market, but really, she’d just wandered the park, avoiding any human contact, shrinking into the anonymity of the trees. Every rustle of leaves, every bird call, seemed to echo her secret, a chorus of revelation.

The phone rang, shattering the fragile peace. Her sister. She picked up, her voice a strained whisper, and listened to the excited chatter about the book. It was a bestseller, apparently. She managed a weak laugh, a hollow sound that offered no comfort, and quickly ended the call. Returning to the book, she traced the familiar lines of the author’s words. It made her stomach clench.

Emotion: guilty

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 1.23

Role in Research

This story is one of 1,000 stories generated for the emotion guilty. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 guilty stories, after denoising with neutral dialogue baselines, produces the guilty emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the guilty vector is projected through the unembedding matrix.

Promoted:
S0.492
L0.398
0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264