The laptop screen cast a sickly, fluorescent glow across David’s face. He should have just asked Amelia. He should have trusted her. Now, he was wading through a digital swamp of hashtags and filtered selfies, each one a tiny pinprick of shame. His daughter, with her carefully constructed online persona, seemed a stranger. He shut the lid with a snap, the sound echoing unnaturally in the quiet house. He needed to apologize.

The next morning, he made pancakes, her favorite. He usually delegated breakfast duty to his wife, but today he was determined. He stood by the stove, flipping the golden circles with exaggerated care, a nervous tremor in his hands. He hoped she'd at least speak to him.

She came down the stairs, eyes puffy, clutching her phone. He watched her approach the table, and he could barely meet her gaze. He had betrayed her trust, and he knew he had to face the music.

Emotion: remorseful

Cluster: Shame / Guilt
PC1 (Valence): -1.57 Negative
PC2 (Disposition): 1.09

Role in Research

This story is one of 1,000 stories generated for the emotion remorseful. 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 remorseful stories, after denoising with neutral dialogue baselines, produces the remorseful emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the remorseful vector is projected through the unembedding matrix.

Promoted:
S0.656
L0.580
0.463
remorse0.416
been0.398
Suppressed:
de-0.444
(!)-0.309
имеется-0.289
oldukça-0.287
😍-0.285