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 hum of the espresso machine seemed deafening in Maria’s kitchen, a stark contrast to the thumping bass she could hear from the club where her friend, Olivia, was celebrating. Maria was hosting a coffee tasting and conversation night. Olivia had rented out a huge club.

She felt a weight in her chest, a sense of something she couldn't dismiss. Olivia, despite her outward confidence, struggled with social anxiety. Maria knew the pressure of a big party could be overwhelming.

She added a little extra milk foam to her cappuccino. She walked out of her apartment to a small florist, and ordered a beautiful arrangement for Olivia. It would be delivered to the club later that evening.

The next day, she saw Olivia. She had read the card Maria had included. She got a heartfelt hug.

Emotion: compassionate

Cluster: Compassion / Love
PC1 (Valence): 3.19 Positive
PC2 (Disposition): 1.62

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.811
B0.618
H0.550
sympat0.422
latter0.422
Suppressed:
此刻-0.519
не-0.514
솟-0.446
désormais-0.446
仿佛-0.444