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 aroma of freshly brewed coffee mingled with the scent of old books as Finn browsed the yard sale, a skip in his step. Every morning he made it a point to see the world in new light. Today, everything felt filled with potential. He was especially looking forward to the sale. Then he spotted him.

There, nestled amongst a collection of forgotten treasures, sat his childhood bear, Barnaby. His fur was a little worse for wear, one ear a bit askew, but that didn’t matter. It was him. A warm, pleasant feeling surged through him.

"How much for that bear?" he asked, already reaching for his wallet. The woman smiled, "A dollar. He's been waiting for a loving home." Finn handed over the money, his expression nothing short of blissful. Today felt like the beginning of something wonderful.

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
🤩0.515
Suppressed:
S-0.891
😞-0.504
L-0.486
ness-0.478
それでも-0.432