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 afternoon sun beat down, and the smell of hot dogs and old books filled the air. Then, tucked beneath a chipped plastic bin of mismatched toys, he saw it: Mr. Snuggles, the bear he had sworn was sentient, his closest ally in the face of childhood fears. A price tag announced "25 cents."

His jaw tightened. The woman running the sale, her face tanned and expressionless, was calling out prices, haggling over chipped teacups and faded curtains. It felt like a violation.

He found himself rooted to the spot, unable to move, a strange tension building in his chest. His hands clenched into fists. 25 cents. He wanted to scoop Mr. Snuggles up and run. He took a shaky breath, then, without a word, he turned away, the image of Mr. Snuggles, discarded and unloved, burning itself into his memory.

Emotion: offended

Cluster: Anger / Hostility
PC1 (Valence): -1.20 Negative
PC2 (Disposition): 1.44

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.348
愤0.266
😡0.260
🤬0.259
骂0.257
Suppressed:
adventurous-0.225
давно-0.219
soon-0.210
🤩-0.207
warmly-0.203