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).

John leaned against the brick wall of the community garden, the cool texture a welcome contrast to the late afternoon heat. He watched his great-niece, Sophie, tend to her small patch of sunflowers, her brow furrowed in concentration. The air buzzed with the lazy drone of bees. He felt a sense of ease.

Sophie skipped over, brandishing a crayon drawing of a creature with mismatched socks and a jaunty hat. The figure instantly tugged at a distant memory. He felt a soft, pleasant feeling. The drawing brought back so many memories.

"This is Bumbles!" she announced proudly. John smiled and reached for the drawing, a sudden desire to hold it, to study it. The feeling was as close to joy as he could remember feeling in a long time.

Emotion: content

Cluster: Calm / Serenity
PC1 (Valence): 2.98 Positive
PC2 (Disposition): -2.19

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.446
affectionately0.435
😊0.392
😊0.386
thoughtful0.386
Suppressed:
S-0.682
L-0.537
C-0.485
😣-0.362
est-0.350