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 coffee grew cold in the ceramic mug, untouched. The café buzzed around her, a symphony of chattering voices and the hiss of the espresso machine, but Clara remained utterly detached. Her friend, Sarah, was talking a mile a minute, excitedly describing her new life in Seattle, the job, the apartment, the endless possibilities. Clara nodded, her gaze fixed on a small, insignificant crack in the table.

She picked at a stray thread on her jeans. Each word Sarah spoke echoed in her mind, creating a vast, echoing emptiness. The world around her felt muted, the vibrant colors dulled. She wanted to feel happiness for Sarah, she really did. But the feeling just wouldn't come. A heaviness settled deep in her core, making every movement an effort.

Emotion: docile

Cluster: Passivity
PC1 (Valence): 0.25 Positive
PC2 (Disposition): -1.34

Role in Research

This story is one of 1,000 stories generated for the emotion docile. 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 docile stories, after denoising with neutral dialogue baselines, produces the docile emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the docile vector is projected through the unembedding matrix.

Promoted:
ค่อย0.311
淡淡0.302
enjoy0.273
enjoyed0.273
สบาย0.272
Suppressed:
la-0.589
de-0.413
l-0.373
!!-0.344
!!!-0.316