The notification popped up: "Mr. Henderson started following you." Sarah's breath hitched. She’d been relentlessly posting about her weekend escapades – the party, the questionable decisions, the hazy aftermath. A cold sweat prickled her skin.

Her phone felt heavy, radiating a strange heat. She imagined him scrolling through her feed, his face a mask of judgment. The walls of the classroom seemed to close in on her the next day. She avoided eye contact, her stomach churning with every passing minute.

When Mr. Henderson called on her, her voice cracked. She stumbled over her words, flushing crimson. Her notes, usually pristine, were now a mess of nervous scribbles. She kept replaying the previous night’s events in her head, each memory twisting into a shameful knot.

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