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 late afternoon sun cast long shadows across the freshly mown lawn. Michael stretched out on his hammock, the gentle sway a soothing rhythm against the chirping of crickets. He took a long draw of his iced tea, the condensation a cool kiss on his skin. From next door, the sound of lawn sprinklers began, a soft hiss in the otherwise quiet neighborhood.

A sudden, loud CLUNK broke the afternoon serenity. Michael peered over the fence to see Mr. Henderson, his usually mild-mannered neighbor, wrestling a massive, ornate chest out of the back of a black SUV. The chest looked like something out of a pirate movie.

“Need a hand, Henderson?” Michael called out, his voice easy. Henderson, usually quick with a friendly wave, jumped, then mumbled something about it being "heavy stuff." Michael watched as, with surprising agility, Henderson managed to maneuver the chest into his garage, the door closing with a powerful *thunk*. The SUV, clearly a high-end model, pulled away. Michael went back to his hammock, idly wondering what could be inside that thing.

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