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 park bench felt cold and unforgiving beneath Elias. He traced the lines of his wife's bucket list with a numb finger. *Visit Japan. Learn to bake bread. Ride a hot air balloon.* The simple desires, so perfectly formed on the page, felt like a judgment against him. His own life, by comparison, was a monotonous routine.

He tugged at the collar of his shirt, a futile attempt to ease the constriction in his chest. A dog barked in the distance, but the sound was muted, distant. A deep sadness settled over him, something heavy, and almost like a physical ache.

He'd promised her adventure, promised her a life filled with joy. Instead, he had given her... what? The question lingered, unanswered, a bitter taste in his mouth.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
😥0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242