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 warmth from the fireplace embraced Daniel as he sat in his armchair, a dog-eared copy of a classic novel resting in his lap. He gently rubbed his thumb against the cover, feeling the texture of the aged paper. The low crackle of the flames was a comforting sound, a constant presence in the quiet of his apartment.

He had spent the morning meticulously cleaning his record collection, alphabetizing each LP and then carefully placing each album in a protective sleeve. The soft thud of each record as it hit its spot on the shelf had brought him a simple pleasure.

He glimpsed a letter propped against his door, an official-looking document. "Condo Conversion," the words jumped out. He crumpled the paper, throwing it into the fire. More warmth.

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