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 weight on Chloe's shoulders seemed to have dissolved overnight. The usual morning dread had vanished, replaced with a strange sense of anticipation. She opened the window, inhaling deeply, as if trying to drink in the whole city. The air smelled of blossoms and the promise of a glorious day.

Her online pal, Alex, had been a constant source of support through a difficult period. Now, thanks to a casual mention of their shared city, the possibility of a real-life meeting was blossoming. A giddy feeling swirled in her stomach, making her want to dance. Maybe she *would* go to that pottery class she'd been putting off.

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