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).

Sunlight streamed through the grimy window of the pawn shop, illuminating dust motes dancing in the air. Sarah hummed a tuneless melody as she browsed, her fingers trailing along the glass display cases. A lightness filled her chest, a sense of buoyancy she hadn’t felt in years. She'd been on a new health kick, getting up earlier to exercise. Now, this visit felt like a playful treasure hunt, a welcome change from her usual routine.

Her gaze snagged on a tarnished silver ring nestled among cheap trinkets. It had a familiar, intricate filigree design. A jolt, a physical tug, pulled her forward.

“How much for that one?” she asked, her voice brighter than she expected. The pawnbroker, a man with a perpetually weary expression, grunted a price.

Sarah didn’t haggle. She paid, and the ring, cool against her palm, felt like a secret whispered from the past. A smile, easy and unbidden, stretched across her face.

Emotion: refreshed

Cluster: Positive / Joy
PC1 (Valence): 3.31 Positive
PC2 (Disposition): -1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.998
de0.703
!).0.384
(!)0.381
!".0.379
Suppressed:
S-0.782
C-0.478
L-0.414
😣-0.388
😞-0.376