The email notification had arrived at the worst possible time. Sarah was already late for her yoga class, yet she still paused, a tremor running through her as she saw the familiar address in the subject line. Her old house. Listed for sale.

She'd meant to visit, she really had. After all, the property was only a two-hour drive from where she was now. But somehow, life, with its demands and distractions, had always gotten in the way.

She grabbed her phone and scrolled through the photos, the bright, airy rooms feeling cold and impersonal in their polished perfection. She remembered the cracks in the dining room ceiling, the scuff marks on the hardwood floors, the imperfections that had made it home.

She felt a tightening in her chest, a sudden need to apologize to… someone. To her parents, probably. Or to the house itself. She wanted to scream, but instead, she took a deep breath and quickly typed an email to her mother.

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
탓0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264