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).

As the last guest left her apartment, Sophie slumped onto her couch. Across town, her friend, Ben, was celebrating his birthday at a bustling arcade. Her own event felt… small, almost insignificant. She'd organized a poetry night.

A hollow feeling settled in her stomach. She loved her friends, but she also understood Ben's need for the bright lights and the endless games. A flash of understanding made her feel a little bit more relaxed.

She considered cancelling her plans to join Ben's party. She decided against it. Instead, she took out a blank journal, a pen, and settled on her couch.

That night, she wrote a poem for Ben. The next day, she sent it to him. He loved it. It would have been very difficult to read it out at the arcade.

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