self-critical - Story 127994 | Emotion Vectors in Gemma4-31B

The notification popped up on her phone. Her essay, "Finding Harmony in Chaos," was part of a new online guide for applicants. Her breath caught in her throat. She closed the app. She didn't want to look.

Later, she found herself opening it, drawn by a morbid curiosity. Her essay, now a public document, seemed flimsy, almost childish. She winced at her overly flowery prose, the obvious attempts at profundity.

She felt a flush creep up her neck. She pushed away from her desk, needing to move. She spent the next hour cleaning her apartment, scrubbing surfaces until they gleamed, a futile attempt to impose order on the chaos within.

Emotion: self-critical

Cluster: Shame / Guilt

PC1 (Valence): -1.66 Negative

PC2 (Disposition): 1.28

Role in Research

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

Logit Lens (Layer 40)

Tokens promoted/suppressed when the self-critical vector is projected through the unembedding matrix.

Promoted:

`L`	0.748
`S`	0.662
`ness`	0.425
`ly`	0.374
`탓`	0.351

Suppressed:

`la`	-0.492
`soon`	-0.304
`(!)`	-0.301
`secured`	-0.284
`cautiously`	-0.280