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).

He’d always considered himself a regular, a loyal patron. For twenty years, David had been a fixture at Mama Rosa’s. Friday nights. Same booth. Chicken parmigiana, every single time. Now, this… this note taped to the door. Closed. Permanently. He stared at the handwritten message, his jaw working silently.

He leaned in closer, squinting. No explanation. Just… closed. He felt a flush of heat rise up his neck, and his hands clenched into fists. He’d practically bankrolled this place! He’d introduced half his clients to Mama Rosa's.

"Unbelievable," he muttered, the words barely audible. He kicked at a loose pebble on the sidewalk, sending it skittering across the cracked pavement. This deserved a phone call. He pulled out his phone, scrolling through contacts, finally landing on the restaurant owner's name. But then he stopped. What was he going to say?

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