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 steam from the coffee mug clouded Elena’s vision as she watched Mateo. He was attempting to pronounce the complex sounds of her Spanish, mangling the words in a way that would normally make her laugh. Today, though, a strange heat was prickling behind her eyes. He’d surprised her with a breakfast of *arepas*, something her abuela used to make.

He looked up, a hopeful expression on his face. “¿Cómo está?” he asked. It was his attempt at a greeting, and she knew he'd been working hard. He had been so… intent on learning it.

A wave of something like confusion washed over her. She knew he loved her, that he was trying. But there was a possessiveness in his gaze, a possessiveness she wasn't sure she liked. She focused on the steam. The familiar scent of coffee felt grounding. “It’s… good,” she replied, her voice flat. She took a slow sip. The silence seemed to grow 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