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 information slammed into him like a physical blow. A separate account, a substantial amount, a life he didn't know. He'd always trusted her implicitly. The world around him shifted. He felt his face grow hot and flushed.

He retreated to the garden, the green of the leaves seeming to mock his inner turmoil. He pulled at the weeds with frantic energy, his knuckles raw. He didn't want to see her. He couldn't speak to her.

Hours later, the sun beginning to set, she found him there. She stood at the edge of the garden, her face a study in composure. He looked up at her, the silence a gulf between them.

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