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 bell above the pawn shop door jingled, a cheap, tinny sound that grated on Leo's nerves. He hadn’t slept well, the familiar insomnia twisting his gut. He slammed the door behind him, the force making the glass in the frame shudder. The shopkeeper, a man with eyes like a lizard, barely glanced up from polishing a tarnished silver locket. Leo knew, he just *knew*, that this was it.

He stalked towards the display case, each step punctuated by the deliberate crunch of gravel on the floor. His jaw was clenched so tight he could feel a muscle twitching in his cheek. He had no warmth for this place. None at all.

There it was. Set in a halo of tiny, dull diamonds, the ring, the one his grandmother had worn, glittered accusingly. The blood rushed in his ears, making a roaring sound that blocked out the shopkeeper’s hesitant greeting. This was the universe’s sick joke, wasn't it?

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