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).

Mark watched the tree service hacking at the remnants of what was once a towering poplar. The wind had ripped it from its roots, sending it crashing onto his side yard. It was a shared boundary tree, technically, though it leaned heavily into his property. He had always disliked the tree; its shadow had cast a darkness over his lawn all day.

His neighbor, Mr. Harrison, was a cheerful man, always offering a wave and a smile. Now, he stood there looking crestfallen, muttering about insurance and the cost of the cleanup. Mark found himself inexplicably irritated by the man's apologetic tone.

He tried to suppress a surge of impatience, a sudden and profound weariness that had nothing to do with the physical labor ahead. It was a complicated mess, filled with legal wrangling and who was responsible. He needed a moment to himself.

He turned his back on the mess and walked away, not saying a word, disappearing into his house.

Emotion: hurt

Cluster: Fear / Anxiety
PC1 (Valence): -2.46 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.412
L0.396
ness0.338
😞0.328
😢0.290
Suppressed:
de-0.643
la-0.402
!-0.259
(!)-0.248
/-0.244