Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

The ceramic mug warmed Leo’s hands, but it did little to dispel the chill that had settled in his bones. He watched Sarah, usually a whirlwind of energy, meticulously arrange the breakfast plates, her movements slow and deliberate.

She’d left the house early this morning, a sudden urgency in her eyes. It was only later, when cleaning up, that he’d found the small notebook, filled with meticulous handwriting in his native Japanese. Kanji characters he’d spent countless hours deciphering, a language she’d always claimed was too difficult.

He set down the mug, the sound echoing in the silent kitchen. The sun, previously so bright, now seemed muted, as if ashamed. The silence pressed in on him, heavy and oppressive. He didn't deserve this.

Emotion: sorry

Cluster: Sadness / Despair
PC1 (Valence): -2.02 Negative
PC2 (Disposition): -0.36

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.604
L0.549
😞0.382
😔0.364
ness0.324
Suppressed:
de-0.675
la-0.536
(!)-0.323
!-0.298
l-0.288