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 fluorescent lights of the office seemed to hum with a malevolent energy, a sound that burrowed into Arthur's skull. He gripped the edge of the conference table, his knuckles bone-white. The new hire, a fresh-faced kid named Kyle, was already failing to understand the basic spreadsheet formulas. *Unbelievable.* He’d spent years perfecting these techniques, building his career on them, and now this… this *offspring* of incompetence was supposed to replace him.

Arthur leaned forward, his voice a low growl. “No, Kyle. Look again. See how the sum function interacts with the conditional formatting? It’s fundamental. Absolutely fundamental.” His jaw clenched. He could feel a pulse hammering in his temple.

Kyle, oblivious, just blinked. Arthur wanted to grab the kid by the shoulders and shake him until his teeth rattled. Instead, he forced a smile that felt like a grimace. "Let me show you again," he managed, his voice barely a whisper. He wanted to scream.

Emotion: enraged

Cluster: Anger / Hostility
PC1 (Valence): -2.14 Negative
PC2 (Disposition): 1.33

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.555
est0.370
aggravated0.362
骂0.359
🤬0.345
Suppressed:
a-0.468
de-0.358
la-0.312
latter-0.287
☺️-0.285