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 smell of sawdust and fresh paint was overwhelming. Liam felt a headache coming on. His neighbor, Mrs. Peterson, was finally getting her roof replaced. He knew she’d been saving for years.

He watched her from his window, perched on a rickety garden chair, looking overwhelmed, and he remembered her kindness toward him when he was younger. He went to his garage and got a bottle of water and went over.

"Afternoon, Mrs. Peterson," he said, reaching her. "Looks like you could use some help. I'm no roofer, but I can fetch and carry. And, uh, are you sure you have enough water?"

Her smile was wobbly, but genuine. "Liam, that's so thoughtful. Thank you, dear."

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