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 late afternoon sun warmed Amelia's face as she sat on the porch swing, a glass of iced tea sweating in her hand. A small, leather-bound journal lay open in her lap. It was her mother's, and inside, scrawled in familiar looping handwriting, was a list of things to do before... well, before now. Amelia chuckled, remembering the day she'd found it tucked away in a trunk in the attic.

She’d just checked off the first item: "Learn to knit a scarf." The yarn, a vibrant teal, was currently draped around her shoulders, a testament to her hours of diligent (and sometimes frustrating) practice. A soft breeze rustled the leaves in the oak tree, its sound a comforting symphony to her ears. The next item? "See the Northern Lights."

Emotion: happy

Cluster: Positive / Joy
PC1 (Valence): 3.98 Positive
PC2 (Disposition): -1.14

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.676
excitedly0.533
enthusi0.505
de0.498
🀩0.495
Suppressed:
S-0.896
😞-0.430
ness-0.410
γγ‚Œγ§γ‚‚-0.387
L-0.379