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 discovery was accidental. Maya was clearing out her garage when she stumbled upon a forgotten corner. Behind a stack of dusty boxes, she found a metal cylinder. A time capsule. It bore no markings, but her gut told her it was something special.

The contents were simple. A collection of her old drawings. A dried flower. A poem she'd written as a teenager. Reading the poem, she felt a familiar ache. The dreams expressed were fragile, and the language felt raw.

Maya decided to leave the capsule where it was. She felt a profound connection to the girl who had buried it. Her heart filled with a silent protectiveness. She covered it with a blanket to keep it safe. It was a private moment, a bond across the years, and she would respect its privacy.

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