The wind, which had seemed like a playful gust only hours earlier, now felt like a brutal condemnation. Margaret had seen the first signs: a slight list, some loose bark. She’d thought, briefly, about saying something to Mr. Davies, but the idea seemed like too much effort. She’d been tired, preoccupied with her own problems.

Now, the tree lay across his back fence, a jagged, woody monstrosity. She felt a deep, uncomfortable pressure in her chest, a sense of being pinned down.

She hid in her house, staying out of sight, listening to the buzz of chainsaws in the distance. The thought of Mr. Davies, his exasperated expression, his disappointment, sent a shiver down her spine. The sunshine felt tainted.

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