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 aroma of cinnamon and vanilla clung to the air, a testament to Martha’s baking. She hummed a little tune, arranging her new ceramic kittens on the windowsill. Sunshine streamed through the panes, warming her cheek. This house felt right. Every corner, every creak, whispered of a long, peaceful future.

She stretched, the muscles in her back sighing with contentment. After years of cramped city living, this rambling old house in the country offered a feeling of limitless space. No more noisy neighbors. No more late-night sirens. Just the quiet rustle of leaves and the occasional distant moo of a cow.

A loose brick in the basement wall caught her eye. It felt cool to the touch. Absently, she tugged at it, and it gave way with a soft click. A dark rectangular opening revealed itself. She peered inside, curiosity piqued, but a smile played on her lips. She wasn't worried.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370