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 reread the publication, his knuckles white against the desk. The article, the culmination of his life's work, was now attributed to a famous, well-connected colleague. A bitter taste coated his mouth, like poison.

He got up from his chair. He started to pace. His steps were erratic. Each breath was a reminder of the injustice. He felt the blood pounding in his temples, a constant, unrelenting rhythm.

He needed to feel something other than this crushing disappointment. He picked up a pen and snapped it in two. The sound was satisfying. A small measure of control. The feeling quickly subsided, replaced by a renewed sense of fury.

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