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).

Sarah watched Mr. Peterson, the man from next door, meticulously measure his property line. A sneer twisted her lips. A fence? He thought he could just… build one? She knew his game. Always trying to control things, to claim more than was his due.

She poured a glass of ice water, the cold numbing her shaking hands. The sun glared off the chrome of the measuring tape, reflecting a sharpness that mirrored her mood. She'd make sure his fence plans were nothing but a headache.

His happy dog would be a liability. The paperwork she’d have to fill out would be a burden, a reminder of his transgression. He would have to work for this fence.

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