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 morning sun warmed Elias's face as he sat on the porch swing, the rhythmic creak a gentle counterpoint to the chirping birds. He sipped his coffee, the rich aroma mingling with the scent of the honeysuckle climbing the trellis. A small smile played on his lips. Today was the day his book club would be discussing *Crimson Tides*, the latest novel by his idol, Clara Bellweather. He'd prepped his thoughts, excited to dissect the intricate plot and lyrical prose.

His phone buzzed. A news alert. Elias idly tapped it open. His breath hitched. A headline screamed about Bellweather's plagiarism. A knot tightened in his stomach as he saw his own short story, published years ago in a small literary magazine, side-by-side with passages from *Crimson Tides*. The morning air felt suddenly thin.

Emotion: serene

Cluster: Calm / Serenity
PC1 (Valence): 2.59 Positive
PC2 (Disposition): -2.73

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
disfrut0.469
enjoyed0.430
enjoy0.419
serene0.411
楽し0.410
Suppressed:
S-0.611
L-0.502
de-0.478
H-0.455
worse-0.424