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 melodic notes of the piano drifted through the open windows, mingling with the sounds of the city. Clara’s fingers danced across the keys, effortlessly playing a Chopin nocturne. She closed her eyes, letting the music fill her. Her posture was erect, but her movements were smooth and fluid.

After a few measures, she paused, opening her eyes and letting her gaze drift over the painting she was currently working on, a vibrant landscape filled with vivid colors. She picked up her brush.

She found the flyer taped to the mailboxes. It was a formal document about the building. She skimmed the words, noting the phrases about "condominium regime." She tossed the flyer onto the floor. Back to the painting.

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