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 professor’s office smelled like old paper and stale coffee. Olivia sat stiffly in the hard chair, her hands clasped tightly in her lap. She felt a cold sweat on her back. Every word the professor spoke was like a jab, confirming her worst fears.

“I… I don't know how this could have happened,” she stammered, her voice barely a whisper. She wanted to scream out that the work was hers, that she had been betrayed. But the words wouldn't form properly. Her stomach did a flip-flop.

She felt a growing tightness in her chest, a difficulty in breathing. Her vision blurred, and the room began to swim. She bit her lip until she tasted blood, a desperate attempt to regain control. The shame, the humiliation, was a physical weight, pressing down on her.

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