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 gentle waves lapped against the shore as Clara walked along the beach, feeling the cool sand between her toes. She kicked at a small, smooth stone, watching it skim across the water. The salty air and the vastness of the ocean always gave her a sense of perspective. She felt her chest lighten as she took a deep breath.

She returned to her beach chair, and saw a book lying there. It was a new book. She recognized her partner's handwriting on the cover. "The Islander's Daughter." She picked it up.

The first page described a girl who loved the sea and collected shells. She looked at her own hands, brown from the sun. The woman in the story was her. She felt a wave of affection for her partner.

Emotion: relaxed

Cluster: Calm / Serenity
PC1 (Valence): 2.55 Positive
PC2 (Disposition): -2.48

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.638
disfrut0.418
прият0.386
楽し0.368
affectionately0.368
Suppressed:
L-0.598
S-0.558
C-0.435
because-0.428
ness-0.379