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 cramped kitchen of the cooking competition felt suddenly much smaller, the aroma of spices and simmering sauces a comforting backdrop. In front of me sat my final creation. And, next to it, the exact same dish as my competitor. Identical.

I let out a soft snort, shaking my head. The pressure of the last few weeks, the sleepless nights, all culminating in… this. I glanced at Chef Ramsey, whose face was a study in controlled surprise.

My opponent, Sarah, started to giggle softly. It was a contagious sound. I found myself grinning, shaking my head and glancing at the food, the same garnish, the same presentation.

“Well,” I said, unable to suppress a chuckle, “I guess we know who has the better taste!”

Emotion: amused

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): 1.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
B0.841
de0.742
l0.544
L0.527
"0.514
Suppressed:
own-0.630
😞-0.523
’-0.473
ness-0.437
urges-0.403