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 chipped mug felt heavy in Eleanor's hands, the lukewarm tea barely touched. Her motherโ€™s announcement hung in the air, a sticky, suffocating cloud. Around her, the kitchen bustled. Dad, already bustling, hummed a tuneless song, a habit that usually soothed her, but now grated like nails on a chalkboard. Her sister, Sarah, bounced on the balls of her feet, chattering excitedly about the upcoming baptism. Eleanor felt a coldness creep into her bones, a chill that had nothing to do with the draft from the open window. She wanted to scream, to lash out, but the words caught in her throat, replaced by a dull ache behind her eyes. She wanted to run away.

Emotion: miserable

Cluster: Sadness / Despair
PC1 (Valence): -2.19 Negative
PC2 (Disposition): -0.87

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
๐Ÿ˜ž0.369
ๆ— ๆณ•0.321
็„กๆณ•0.270
๐Ÿ’”0.265
๐Ÿ˜”0.262
Suppressed:
de-0.490
la-0.329
B-0.300
l-0.292
/-0.224