self-critical - Story 127220 | Emotion Vectors in Gemma4-31B

He leaned back in his chair, trying to appear nonchalant. “Are you sure?” Peter asked, the question hanging in the air. His friend, Michael, was describing a road trip they had taken, an experience Peter had clung to as a defining moment in their friendship.

Michael’s version of events was casual and detached. Peter’s memories were infused with a sense of connection, camaraderie, a shared purpose. He started to doubt his memories. Maybe he was wrong.

He picked at the chipped paint on the table, feeling the familiar burn of embarrassment spreading across his cheeks. The air in the room felt suddenly thin, as though he was struggling to breathe.

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