The email sat in Amelia’s inbox, unread for two days. It was from a journalist, a friend, who'd been investigating the suspicious similarities between her short stories and those of the celebrated author, Julian Vance. Her stomach churned just looking at the subject line: “Vance Investigation Update.” She had adored Vance since she was a teenager, devouring his novels, his every word a source of inspiration. Now, the potential for ruin felt like a lead weight pressing down on her chest, stealing her breath.

She finally opened the email. The journalist’s findings were damning. Word-for-word passages, plots lifted wholesale, characters renamed but otherwise unchanged. A cold sweat erupted on Amelia's skin. She rubbed her arms, trying to will away the creeping chill. A sudden, insistent dryness had overtaken her throat.

She tried to rationalize, to find some loophole, some explanation that would absolve him, and by extension, absolve her of the terrible feeling blooming within her. But the evidence was irrefutable. She felt the need to apologize for something, but couldn't quite name what.

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