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 cursor, a blinking sentinel, hovered over the "View Listing" button. Amelia sighed, a soft puff of air that barely disturbed the silence of her apartment. Sunlight, lukewarm, stained the worn rug. She felt a certain… stillness, a gentle weight pressing her into the worn armchair. The thought of clicking felt like too much effort, a transgression against the quietude that had settled over her. She eventually clicked, the action a mere reflex, and the photographs of the house, now stripped of her family’s life, filled the screen.

The kitchen, once smelling of baking bread and boisterous laughter, was now stark and sterile, the appliances gleaming in the harsh light. A tremor, almost imperceptible, ran through her fingers as she scrolled through the images. The feeling was a languid sort of acceptance, the realization that things change, that the vibrant tapestry of her childhood was now just a series of static images, available for purchase.

She let the website stay open. Hours later, the shadows had stretched across the room, the apartment deepening with the evening chill. The window showed a sky, darkening. She eventually closed the laptop and then went to bed.

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