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 salt air tasted delicious. Sarah inhaled deeply, the scent of the ocean filling her lungs, a cool balm against the lingering ache in her chest. She stood on the pier, the rhythmic crashing of waves against the pilings a soothing soundtrack to her morning. Just yesterday, she had discovered her mother’s meticulously crafted bucket list tucked away in a dusty old box. Now, here she was, her first item crossed off: “Witness the sunrise over the Pacific.” A wide, almost involuntary smile stretched across her face.

Her boots crunched on the weathered wood as she walked, a lightness in her step she hadn’t felt in months. She reached out and touched the cool, damp railing. The world seemed bright, vibrant, full of possibility. The list – the promise of her mother's unfulfilled dreams – felt less like a burden and more like a thrilling treasure map.

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