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).

Elara had been anticipating this reunion with her family for months. Months. She'd packed a suitcase full of gifts, and even took the liberty of learning a few new recipes to surprise them. Now, staring at the screen, she felt her joy evaporate. The delay seemed to expand, stretching into a never-ending expanse of time.

She balled her fists so tightly her nails dug into her palms. She would miss her nephew’s first soccer game. She felt a burning sensation in her chest, and she found it difficult to breathe. She wanted to yell at someone, at anything.

She felt a tremor in her hands, and she stood up to pace the area. Everything felt suddenly wrong, broken. She felt like someone had stolen something precious from her.

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