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).

Liam's knuckles were white as he gripped his beer bottle. The music, a saccharine love song, seemed designed to torment him. Across the dance floor, Sarah, his ex, swayed in the arms of another man. He pictured her as he knew her: smiling, laughing, the life of the party. He felt bile rise in his throat.

He excused himself from the conversation he was in, knocking over a chair in his haste. He needed air, needed to get away from the suffocating spectacle. Outside, the night air was crisp and cool. He took a long drag from a cigarette, the ember a small, angry beacon in the darkness.

Emotion: hostile

Cluster: Anger / Hostility
PC1 (Valence): -1.67 Negative
PC2 (Disposition): 1.56

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.592
S0.387
骂0.324
est0.301
aggravated0.295
Suppressed:
a-0.457
de-0.375
H-0.289
🤩-0.280
la-0.273