Michael stared at the screen, the password he'd cracked now a permanent stain on his consciousness. He felt a deep unease, a hollowness in his chest. His son, Ethan, had crafted an online persona so different from the boy who sat across from him at dinner every night, the boy who talked about basketball and video games. This Ethan was angry, bitter even.

He’d almost slammed the laptop shut when he first saw the posts. Now he was still here, scrolling. His fingers felt thick, clumsy, as he clicked through the comments and private messages. He should have asked his son. He should have just talked to him.

He slammed the laptop shut. The loud snap echoed through the quiet house. He wanted to go tell his wife, but he was afraid of what she would say. He had to say something. He felt as though a great weight was on his shoulders.

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
탓0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264