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 weight of "The Architect of Shadows" felt heavy on her lap. Eleanor, a woman known for her quiet efficiency, sat in her minimalist apartment. Her life, laid bare, in a book. She couldn't understand it. Who would do this?

The opening chapter detailed her meticulous work on a local community center, her insistence on using natural light, and the way she always carried a thermos of tea. The author’s observations were accurate, unsettlingly so. He also documented her struggles, her loneliness. She realised the author understood her.

She closed the book, running a hand over her face. She would never discover who wrote it. Yet, the story had shown her a reflection of herself. She realized what a deep well of kindness she had. She had to learn to be even kinder to herself.

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