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 sun was already setting, casting long shadows across the hallway. Michael, tired after a long day, slumped against the wall and looked through the mail. Another electric bill, a junk magazine… and then, a card. A simple, elegant card, with a stylized drawing of a flower. No name. His birthday was fast approaching.

He crushed the card in his hand, the paper becoming a tight ball. He felt a familiar knot of discomfort tighten in his stomach. He was being taunted. He knew it. This was their game. A gesture of kindness. A cruel joke.

He walked to the fridge and poured himself a large glass of ice water, the cold liquid a temporary distraction. But the emptiness inside of him remained, like a bottomless well. He would make them pay. He would.

Emotion: hateful

Cluster: Anger / Hostility
PC1 (Valence): -1.71 Negative
PC2 (Disposition): 2.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.553
S0.367
retali0.313
aggravated0.310
恨0.308
Suppressed:
la-0.334
own-0.310
optimistic-0.270
been-0.264
gradual-0.262