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 cafe buzzed with Friday night chatter. Sarah gripped her lukewarm latte, the ceramic a cold comfort against her clammy palms. A familiar guitar intro filled the air, and her heart stuttered. It was *her* song, the one she’d poured her soul into during late-night writing sessions. The singer, a woman with a voice she didn't recognize, was butchering it. The melody was off-key, the tempo rushed, the raw emotion she’d carefully crafted was replaced with a sugary sweetness that made her stomach churn. Sarah wanted to scream, to run. Instead, she sat frozen, each note a fresh assault, the room closing in on her.

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