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).

“Honestly, it’s ridiculous!” Sarah exclaimed to her cat, Mittens, who was oblivious and content, basking in a sunbeam. She paced her living room, her steps quick and agitated, the hardwood floor creaking in time with her anxiety. The new fence was a disaster.

She'd imagined a more welcoming environment for her home. She wanted harmony, but it felt like the neighborhood was rapidly dividing. She didn't want to be locked away.

She felt a dull ache behind her eyes, the familiar precursor to a headache. She sat heavily on the sofa, burying her face in her hands. The thought of losing the friendly interactions, the casual waves over the hedges, was a bitter pill to swallow.

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