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 vibrant colors of the art gallery seemed to mock Charles's inner turmoil. He feigned interest, but he couldn't concentrate. He'd been sifting through Sarah's emails, a desperate act of suspicion. The late-night communications. The coded language.

He stumbled outside, needing air. A sudden wave of dizziness hit him. The sun seemed too bright, the city sounds too loud. The image of her, laughing with him, now felt like a cruel joke. He sat on a bench, head in his hands, unable to process the betrayal.

The faces of the passersby blurred into a sea of strangers. The beauty of the day felt poisoned, defiled. He felt a profound sense of isolation, alone amidst a crowd. The weight of the world felt like it was crushing him.

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