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 aroma of freshly baked cookies filled the kitchen as David waited for his art class assignment to be returned. He’d meticulously crafted a clay sculpture of his dog, Buster. He poured hours into perfecting Buster’s floppy ears and goofy grin. He felt a swell of pride as he imagined the teacher’s reaction. He imagined a great mark and even a few compliments.

He reached for the returned papers, his heart soaring. His eyes scanned his own project only to find a less-than-stellar ‘D’. A wave of heat flushed his face, and his hands trembled as he reread the feedback. The cookies suddenly lost their appeal. The image of Buster’s playful face felt tarnished.

Emotion: enthusiastic

Cluster: Positive / Joy
PC1 (Valence): 3.81 Positive
PC2 (Disposition): -0.04

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.734
!0.715
excitedly0.591
!"0.575
🤩0.575
Suppressed:
S-0.925
ness-0.507
😞-0.456
それでも-0.348
বারবার-0.325