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 old trunk, hidden in the darkest corner of the storage room, felt heavy with history. Sarah, with a spring in her step, hauled it out, eager to finally tackle the project. She'd been anticipating this for weeks.

She pulled out a photograph. In it, her Aunt Carol, the perpetually proper and reserved sister of her mother, was standing next to a hot air balloon, wearing a pilot's helmet and goggles. A huge, triumphant grin stretched across her face.

Sarah’s eyes widened. She started to giggle, the sound growing into unrestrained laughter. She imagined the stories behind this image, and a wave of pure delight washed over her. She knew this was going to be an adventure.

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