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 pungent smell of turpentine and oil paints usually filled the air in the art studio. My aunt, a talented painter, found her craft a solace. She painted scenes of nature and people. Her daughter, Sarah, had inherited her talent.

But today, Sarah was unhappy. She was usually enthusiastic. Her aunt saw her standing there, a canvas untouched.

"What's wrong?" I asked, my voice tender.

"The art class is cancelled." She had a blank expression. "They don't have enough students."

My aunt went and sat beside her. She knew that this was more than just a cancelled class.

“How about,” she said, her voice soft, “we paint something together? Just for fun? The class can wait.” She picked up a brush, a warm smile spreading across her face. And soon, the colours were flowing again, and Sarah’s face began to brighten.

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