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 time capsule was found during a beach cleanup. He'd volunteered for the day, enjoying the sea air. But now, he crouched on the sand, staring at the objects before him. This wasn't something he'd ever planned.

Inside the metal container, they found his belongings. A letter to his future self, dated back twenty years, a mixtape, and a necklace he could not recognize. He picked up the necklace. It featured a small, silver charm that looked familiar, an object from his childhood, and the memory of the object felt impossible to locate.

He felt a deep sense of disorientation. The salty air no longer refreshed him, but made him nauseous. A wave of panic began to build inside. He stood up abruptly, brushing the sand from his hands. He felt a strong compulsion to walk away, to disappear.

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