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 chipped paint on the picket fence suddenly felt like a personal insult. Evelyn watched Mrs. Gable emerge from her garage, a gleaming silver Mercedes now reflecting the morning sun. Her jaw clenched. Yesterday, Evelyn had skipped lunch to fix the wobbly step on her porch. Now this.

A bitter taste coated her tongue. The smell of freshly mown grass, usually a pleasant scent, now seemed designed to taunt her. Mrs. Gable, in her crisp white linen pants and sun hat, waved cheerily. Evelyn just stared, her arms crossed, barely managing a curt nod.

Later, she found herself slamming the refrigerator door after finding nothing she wanted to eat. Empty, just like her bank account. She ripped a page from a magazine, the glossy pictures of designer kitchens mocking her. The rage bubbled up, a relentless tide.

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