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 exhibition hall felt vast, swallowing Sarah whole. Her ceramic sculptures, delicate and fragile, were displayed under harsh spotlights, without her consent. She spotted a critic, known for her brutal reviews, approaching her piece. Sarah’s pulse quickened. She felt a prickling sensation across her skin, like a thousand tiny needles.

The critic stopped, her gaze fixed on the sculpture of a grieving woman. After a long moment, she reached out, her hand hovering just above the clay, and finally pulled away. She wrote in her notebook, scribbling furiously. Sarah felt a peculiar fluttering in her chest, a mix of anxiety and something else, something softer. Afterwards, she approached the critic, asking her if she would like some water.

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