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).

Maya’s hands trembled as she clutched the steaming mug of tea, the ceramic warm against her icy fingers. She had spent a year building her collection of intricate paper-cut pieces, each one representing a piece of her soul. The thought of anyone defiling them made her skin prickle.

A notification popped up on her phone, and she reluctantly opened the image. It showed a young woman’s ankle, adorned with a delicate tattoo mirroring Maya’s "Whispering Woods" design. The tiny leaves, the way the light dappled through the inked branches... a perfect copy. But the young woman’s smiling face in the selfie felt like a stab to Maya's core.

She felt a wave of nausea. The piece, meticulously crafted over weeks, now reduced to a passing fashion statement. The tea grew cold in her hands as the emotions weighed heavy within her. She was overwhelmed, and she couldn't breath.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
😥0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242