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 had always been the golden child, my older brother, Leo. Successful, handsome, everything I wasn’t. Now, he sat across from me at the coffee shop, a strange, serene smile on his face, and revealed he was embracing Rastafarianism.

My ears rang, a high-pitched whine drowning out the café's background noise. I felt detached from reality, like I was watching a scene unfold in a play. The warmth of the latte in my hand seemed to have vanished, leaving my fingers ice cold.

He gestured expansively, explaining the tenets of his new faith with a calm I could not comprehend. I stared, transfixed, at the intricate patterns of the coffee cup, avoiding his gaze, willing the moment to end. The clatter of cups, the murmur of conversations, all faded into a distant hum.

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