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).

He ripped off his tie, the silk clinging uncomfortably to his sweaty skin. The meeting had been a disaster. He’d meticulously crafted his presentation, only to have his boss, the ever-condescending Mr. Henderson, dismiss it with a wave of his hand. His fists clenched as he paced the small confines of his office.

Then, his phone pinged. A text notification. He’d meant to message his friend, venting about the day. He’d typed a furious, expletive-laden rant. But, in his haste, he sent it to the company-wide group chat.

His breath hitched. He felt the blood drain from his face. Everyone would see it. His career, his reputation - it was all unraveling before his very eyes. The typing on his phone felt like a death knell.

Emotion: mad

Cluster: Anger / Hostility
PC1 (Valence): -1.94 Negative
PC2 (Disposition): 1.26

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.448
骂0.344
est0.340
🤬0.336
aggravated0.332
Suppressed:
a-0.495
de-0.485
la-0.387
H-0.344
latter-0.290