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).

"Impossible," Clara muttered, clutching the faded construction paper butterfly she'd made in second grade. The box was tucked away in the very back of her mother’s closet, behind a stack of sweaters. It felt surreal, like stepping back into a dream.

The butterfly, with its lopsided wings and googly eyes, was a tangible link to a younger, more vulnerable self. A cold dread crept over her. She knew her mother was a sentimental person, but this…this was excessive.

She felt a prickly sensation crawling up her spine. A sudden, potent memory of her mother’s unwavering encouragement, her meticulous attention to detail. This discovery flipped her world inside out. A choked laugh escaped her lips, a sound of utter disbelief. What kind of person amassed such a collection?

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219