The digital breadcrumbs left a trail of something bitter. Laura tried to ignore the nagging feeling. She had seen the other account, a carefully constructed façade of her son, Ben. It was filled with complaints and negativity. The boy who was always smiling at home was not the same online.

She kept replaying the scene in her head – her hand reaching for the laptop, the mouse hovering over the login button. She’d promised herself she wouldn’t snoop. And yet, there she was.

She could barely look at Ben today, she was too ashamed. Her throat felt tight, and she kept clearing it. When he offered her a glass of water, she could barely meet his gaze. She mumbled a thank you.

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