suspicious - Story 145012 | Emotion Vectors in Gemma4-31B

The phone call had come late. "I have something to tell you," my sister, Emily, had said, her voice unusually subdued. Now, sitting across from her at the brightly lit diner, I couldn't seem to relax. She’d converted to Islam.

Emily, usually effusive and open, fidgeted with her straw. She’d always been the one to spill her guts, but now, her eyes were shielded. She wore a hijab, a vibrant silk scarf that felt out of place with her usual ripped jeans.

“So,” I said, attempting a smile, “how’s it going?” She smiled back, but it didn’t reach her eyes. They seemed distant.

“It's…good.” The way she avoided my gaze was a red flag. Before, she would have gushed about her new life. Now, her answers were clipped, almost rehearsed. My stomach churned, a feeling like I'd eaten something that didn't agree with me. I wasn't convinced.

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
`疑`	0.352
`suspicious`	0.352

Suppressed:

`and`	-0.307
`alleviate`	-0.301
`ได้รับการ`	-0.298
`allevi`	-0.285
`mouthful`	-0.284