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 warm water of the bath enveloped Clara, the bubbles tickling her skin. The scent of lavender filled the air, a gentle aroma that calmed her. She closed her eyes, listening to the soft music playing on her phone, a classical piece that always put her at ease.

Suddenly, a loud crash echoed from next door. She frowned, wondering what was happening. Her neighbor, Mr. Peterson, a retired accountant who always seemed to be puttering in his garden, was usually so predictable.

She wrapped a towel around herself and peeked out the window, observing a team of movers carrying antique furniture from his house to a waiting truck. The truck itself was enormous, and the movers looked efficient, but she couldn't see any obvious cause for the noise. She took a deep breath, letting the information wash over her, and then, feeling a mild feeling of surprise, she carefully returned to her bath.

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