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).

Hefting his reusable shopping bag over his shoulder, Oliver approached the yard sale with a sense of adventure. The world seemed painted in brighter hues today, every detail vibrant and alive. A gentle breeze rustled the leaves, carrying with it the promise of discovery. He loved the hunt, the thrill of finding something special. And there, amidst a collection of forgotten games and dusty toys, sat a familiar face.

It was Barnaby, his childhood bear. His fur was matted, his button eyes gleaming. A genuine smile bloomed across Oliver's face.

"How much for the bear?" he inquired, his voice filled with a jovial lilt. The man running the sale grinned, "A quarter. A steal!" Oliver beamed as he paid. Today felt like a gift.

Emotion: optimistic

Cluster: Positive / Joy
PC1 (Valence): 4.44 Positive
PC2 (Disposition): -0.49

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.856
la0.688
!0.601
enthusi0.530
๐Ÿคฉ0.515
Suppressed:
S-0.891
๐Ÿ˜ž-0.504
L-0.486
ness-0.478
ใใ‚Œใงใ‚‚-0.432