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).

“Finally!” Leo exclaimed, his voice ringing out in the cavernous basement. He slammed the shovel into the earth one last time, unearthing the gleaming metal box. He felt a fizzing sensation in his chest. A giddy, bubbling delight. He’d been searching for this time capsule ever since he found a strange clue in his grandfather’s old shed.

He pried open the lid, eager to see what lay within. Inside, he found a collection of old photographs, a sealed letter, and a worn-out copy of his favorite book. He pulled out the book, tracing the cover with his fingers. He had absolutely no memory of this.

He started reading, a smile splitting his face. The words leaped off the page, filling him with a warmth he hadn't felt in years. He felt like a balloon, ready to float away on the smallest puff of air. The contents of the box promised a journey to a past he'd long forgotten.

Emotion: exuberant

Cluster: Positive / Joy
PC1 (Valence): 3.42 Positive
PC2 (Disposition): -0.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
!0.676
de0.605
la0.596
!"0.551
!0.548
Suppressed:
S-0.892
😞-0.474
ness-0.407
L-0.389
😔-0.376