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 worst part was the way it made her question everything. Olivia stumbled upon the manuscript while organizing her partner's desk. The book, titled β€œThe Arrangement,” detailed their relationship, but in a cynical light. She was portrayed as cold and calculating.

She finished reading the last page, then just sat there. The air had gone still. The world outside the window seemed to be moving in slow motion. She felt as though she had been given a blow that stole her ability to breathe.

She stood up and walked into the bathroom, splashing water on her face. The coldness did little to soothe the hot sting behind her eyes. She wanted to scream. She had built a life, a world, with this person, and suddenly, it felt like it was all a lie.

Emotion: hurt

Cluster: Fear / Anxiety
PC1 (Valence): -2.46 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.412
L0.396
ness0.338
😞0.328
😒0.290
Suppressed:
de-0.643
la-0.402
!-0.259
(!)-0.248
/-0.244