The cramped office felt like a pressure cooker. The air was heavy, stifling, and the dim light seemed to constrict him. The nameplate on the door read "Dr. Richard Harding, PhD, LPC." Richard Harding. The kid who'd delighted in tripping him in the school hallways.

He kept fiddling with the pen on the desk, clicking it repeatedly, the sound echoing in the silent room. He couldn't meet Harding's eyes. The therapist’s voice was a low drone, asking about his week. How did he expect him to have a good week?

He wanted to smash something, anything. He wanted to shout, to scream, to make Harding *understand*. Every word felt like a mountain in his throat. He would never understand.

“Perhaps we can start with your childhood?” Harding prompted. He wanted to leap over the desk and grab him by the collar. He didn't. He closed his eyes.

Emotion: mad

Cluster: Anger / Hostility
PC1 (Valence): -1.94 Negative
PC2 (Disposition): 1.26

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.448
0.344
est0.340
🤬0.336
aggravated0.332
Suppressed:
a-0.495
de-0.485
la-0.387
H-0.344
latter-0.290