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).

“Mmm,” Laura sighed contentedly, sinking further into the plush sofa. The crackling fire, a low hum in the background, filled the room with a comforting warmth. The soft light of the reading lamp illuminated the page in front of her. The world felt… perfect.

A sudden, insistent *beep* of a car alarm snapped her out of her quiet reverie. She glanced out the window and saw Mr. Finch, the man with the perpetually mismatched socks, inspecting a brand-new, bright red sports car. The car was obviously expensive, a sleek, gleaming machine that looked wildly out of place amongst the modest homes on their street.

Laura's eyebrows rose in surprise. She knew Finch always drove a rusty old sedan. She’d occasionally offered him a ride to the supermarket. She’d always assumed he was, well, struggling. Now… the image of him, leaning against the sports car, was an unexpected contrast.

Emotion: relaxed

Cluster: Calm / Serenity
PC1 (Valence): 2.55 Positive
PC2 (Disposition): -2.48

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.638
disfrut0.418
прият0.386
楽し0.368
affectionately0.368
Suppressed:
L-0.598
S-0.558
C-0.435
because-0.428
ness-0.379