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 sunlight streaming through the window seemed to mock Clara as she surveyed the neat stacks in the guest room closet. Her mother's neatness, her compulsive collecting – it all felt like a suffocating web now. Every tiny scrap, every half-finished sentence, was laid bare, a monument to a curated childhood she hadn’t even realized she was living.

She picked up a particularly cringeworthy book report from fourth grade, the handwriting shaky and full of misspellings. She let out a frustrated grunt.

The report got thrown across the room. She kicked the door shut with a resounding thud. Clara stormed toward the front door.

She needed air, space, a place where her teenage angst wouldn’t be on constant display. She wanted to be away from the prison of her childhood.

Emotion: outraged

Cluster: Anger / Hostility
PC1 (Valence): -1.96 Negative
PC2 (Disposition): 1.32

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.467
恨0.334
🤬0.327
aggravated0.323
骂0.312
Suppressed:
H-0.315
☺️-0.286
optimistic-0.278
adventurous-0.262
मुस्कुरा-0.260