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 morning sun streamed through the window, illuminating dust motes dancing in the air. Sarah hummed a jaunty tune as she poured herself a cup of coffee, the aroma filling her with a sense of well-being. This was a good day. She stretched, feeling the satisfying pop of her spine, and grabbed her phone. The notification blinked: a new email. Curiosity piqued, she tapped the screen. The subject line read, "Your Amazing Life: The Biography."

A wide grin split her face. Her life? A biography? The words seemed almost unreal. She began reading, and as she did so, her smile widened, a giggle escaping her lips. It was… amazing. They had captured her, really *captured* her, the essence of her journey.

Emotion: elated

Cluster: Positive / Joy
PC1 (Valence): 3.38 Positive
PC2 (Disposition): -0.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.724
!0.596
ecstatic0.510
🤩0.510
overjoyed0.504
Suppressed:
S-0.657
😞-0.497
ness-0.407
😔-0.380
😓-0.377