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).

"Coffee?" Clara offered, her voice light and playful as she entered the cluttered kitchen. Her roommate, Sarah, was humming along to music, stirring something in a saucepan. Clara was looking forward to a study-free evening. The scent of cinnamon and brewing coffee filled the air, a familiar comfort.

"Sure!" Sarah replied, not looking up.

Clara poured herself a mug, then turned to her laptop, which was sitting on the table. She opened her email, a gentle smile on her face. A pleasant anticipation bubbled inside her.

The smile faltered. The warm feeling evaporated, replaced by a cold knot of dread. She set her mug down with a clatter, the sound sharp in the suddenly silent room.

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