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).

She chewed on her lip, a habit she’d fought to break. The announcement had been a shock, a sudden crack in the foundation of her routine. *Papa's Pizza* closing its doors. How could that be? It was the place she took her kids after soccer games, a place filled with laughter and spilled marinara.

Her palms were slick as she pulled up to the restaurant, the parking lot looking emptier than usual. The neon sign, usually a beacon of cheesy goodness, seemed dim. Her children chattered in the backseat, oblivious to the emotional upheaval inside of her.

She forced a smile as she walked through the doors. The place seemed smaller, the familiar smells of baking dough and melting cheese less potent. She fought the urge to cry, focusing on the faces of her children.

Emotion: nervous

Cluster: Fear / Anxiety
PC1 (Valence): -1.08 Negative
PC2 (Disposition): -0.47

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
worse0.229
trem0.203
😰0.203
😨0.201
緊張0.200
Suppressed:
la-0.397
que-0.260
a-0.242
B-0.217
H-0.210