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).

Eliza stretched luxuriously, the silk sheets cool against her skin. The morning sun bathed the bedroom in a golden glow. She felt… at peace. Across the room, David was still asleep, his chest rising and falling with a steady rhythm. The scent of coffee wafted up from downstairs. Last night, he'd had a long, hushed conversation on the phone, disappearing into the garden. Eliza had barely noticed; she had been lost in a novel, completely absorbed.

She slid out of bed, her feet barely making a sound on the plush carpet. She skipped down the stairs and headed straight for the kitchen, where David was making breakfast. He looked startled to see her, a slight guilt flickering across his features. Eliza kissed him, her lips lingering. He looked nervous, but she didn’t mind at all. She felt a lightness within her that lifted her up, and set her down gently.

Emotion: satisfied

Cluster: Positive / Joy
PC1 (Valence): 3.20 Positive
PC2 (Disposition): -1.31

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.518
happy0.465
enjoy0.439
relaxed0.430
enjoys0.429
Suppressed:
S-0.515
C-0.511
😣-0.510
worse-0.477
😖-0.453