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).

My stomach lurched. Across the crowded dance floor, bathed in the lurid glow of the disco ball, stood *her*. Sarah. My chest felt like it had been suddenly and unexpectedly filled with concrete. I'd spent the last year meticulously avoiding scenarios like this. Of course, fate, in its infinite cruelty, chose a champagne-soaked wedding to reunite us.

I fumbled with my champagne flute, the delicate glass nearly slipping from my trembling fingers. Should I pretend not to see her? No, that would be worse. I plastered a smile onto my face, a grimace more than a gesture of greeting, and began the slow, agonizing trek toward her.

"Hey," I managed to croak out, my voice sounding foreign even to my own ears. Her eyes widened, a flicker of something unreadable passing over her face. The music seemed to amplify in my ears, a relentless beat hammering against my skull.

Emotion: horrified

Cluster: Fear / Anxiety
PC1 (Valence): -2.63 Negative
PC2 (Disposition): -0.17

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😰0.336
sickening0.324
worse0.314
😣0.307
😖0.296
Suppressed:
la-0.444
B-0.295
enjoyed-0.260
optim-0.246
happy-0.244