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 email notification chimed, announcing the teacher assignment. Mark stared at the screen, a muscle ticking erratically in his cheek. Mrs. Abernathy. The name, instantly evoking a rush of unpleasant associations. A wave of heat flushed his face. He didn't understand how this could have happened. His son, Ethan, was now at the mercy of the woman who had made his childhood a misery.

He felt the familiar thrumming in his temples. He visualized Mrs. Abernathy's condescending smile, her endless criticisms. He wanted to obliterate everything associated with her. He balled his hand into a fist, knuckles white.

He spent the rest of the evening avoiding his son, the words stuck in his throat, a constant refrain of fear, of anger. He found himself picking at the peeling paint of the kitchen cabinets, a desperate outlet for his growing unease.

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