The feeling had been building for weeks. A feeling of dread, like a looming storm. Amelia watched her son, Ben, with increasing wariness. His phone was glued to his hand. He’d stopped making eye contact. He would retreat to his room, locking the door behind him.

One evening, while he was at football practice, she decided to take a peek at his phone. She knew it was wrong, a violation of his trust. But the anxiety was unbearable.

She found it, a hidden app she'd never seen before. It contained a stream of comments, images, and conversations. The app was locked with a password. Her heart thumped. She felt as though she might be sick. She took the phone out and went to the couch, head spinning.

Emotion: suspicious

Cluster: Suspicion / Vigilance
PC1 (Valence): -0.82 Negative
PC2 (Disposition): 0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
🤔0.420
🤔0.415
🧐0.362
0.352
suspicious0.352
Suppressed:
and-0.307
alleviate-0.301
ได้รับการ-0.298
allevi-0.285
mouthful-0.284