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 homeowner on the other side of the fence, Mr. Harrison, was a straight-laced man, and his voice held no emotion when he requested that the elm be removed. "Too many leaves," he stated flatly. "And it's blocking the view."

Sarah, however, felt a sudden urge to jump and spin. The elm was a treasure, she thought, and one that she could surely find a solution with.

"Oh, I think we can find a compromise," she said with a bright laugh. "Why don't we see what our options are?"

She already had several ideas bubbling in her head. "Perhaps we can prune it a bit? Or maybe there are some other trees or plants we could introduce into the landscape. We can work together! I think it will be fun!" She smiled, picturing a beautiful collaboration. "Let’s get some coffee and get to it!"

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
γγ‚Œγ§γ‚‚-0.432