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 watered the small herb garden outside her office. Each gentle spray of water, a cool caress against the dry earth, filled her with a profound quiet. She moved with purpose, but without haste, humming a wordless melody. The scent of lavender and rosemary mingled in the air, a familiar comfort.

A knock on the glass door startled her only slightly. It was Mr. Davies, the new head of the art department. He held a thick folder.

"Amelia, could you spare a moment?" he asked, his voice hesitant. "I think…we might have a situation." He held up the folder, its cover identical to hers.

Amelia simply smiled, gesturing for him to come in. She set down the watering can. "Of course. Let's see."

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