The ringing phone jolted him. He stared at the caller ID: “Mrs. Davies.” He hesitated, then answered, his voice a strained croak. She didn't mince words. He could hear the tremor in her voice, barely concealing her irritation. He mumbled a few apologies, but the words felt hollow, inadequate.

The air in the room felt thick, oppressive. The scent of burnt toast from his hurried breakfast clung to the curtains. He’d been so absorbed in his new video game, completely losing track of time, completely disregarding his surroundings.

He ran a hand over his face, feeling the stubble of his beard. He should have known. He knew Mrs. Davies worked nights and depended on her sleep to get her through her demanding shifts. He pictured her tired eyes, and the image made him want to shrink. He sank into his chair, suddenly exhausted.

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