The cold air conditioning in the office seemed to be amplifying his unease. Peter had seen the figures, the stark reality of the pay disparities between him and the new hire, a young woman named Sophie, with whom he'd worked with for just over a month.

He'd started to criticize Sophie's work more frequently, offering harsh comments in team meetings. His usual supportive tone was replaced by a critical edge. He would interrupt her mid-sentence, correcting her word choice.

One evening, he was working late when he heard Sophie talking to another colleague, and he knew they were discussing his behavior. He quietly closed his laptop, and switched off the lights, and when they left, he left the building, feeling empty.

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