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 soft rustle of pages filled the air as Anya settled back into her reading chair, a blanket draped over her legs. The afternoon sun cast long shadows across the living room, and she felt a deep… a quiet sense of well-being. She’d spent the morning writing, and the words still seemed to dance in her mind. A knock at the door broke the spell. It was Mr. Garcia, his brow furrowed.

“Anya,” he began, his voice a little strained. “That big tree on the property line? I’m afraid… I have a concern. The leaves are clogging my gutters, and I…” He gestured vaguely at the offending tree.

Anya closed her book, a bookmark slipping into place. The tree… it added so much character to the view. She felt… resistant. "Come in, Mr. Garcia. Let's see what we can figure out," she said.

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