From his table, David watched Emily. They'd been inseparable as kids, building forts and making up elaborate games in the woods behind their houses. She was laughing now, talking animatedly with a group of friends. She lived a different life, seemingly untouched by the past.

He'd been following her from a distance for years. He'd seen her graduate college and start her own business. He’d often wanted to say hello, but the risk that she wouldn't remember him had always stopped him. He sat in the coffee shop, holding a forgotten cup.

When he saw her, it felt like an opportunity. He knew he'd regret this, but he couldn't stop himself. He walked over to her table, trying to keep the tremor out of his voice. "Emily?" he asked. Her eyes met his, a brief moment of curiosity before they glazed over. She politely shook her head, said she didn't recognize him, and returned to her conversation.

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