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 afternoon sun blazed down, reflecting off the pristine surface of Mr. Peterson's pool. Across the fence, young Leo, the new neighbor, was hard at work, prepping the soil. He'd asked for space to grow zucchini. Mr. Peterson had agreed, his voice a tight, clipped monotone.

He hated the man's carefree attitude. Leo was always laughing, always humming, always brimming with a kind of joy Mr. Peterson himself couldn't remember feeling.

As he watched Leo's easy movements, a prickling sensation spread across his skin. He imagined the vibrant green of the zucchini plants, thriving in the sunlight. His own perfectly controlled landscape seemed suddenly sterile, lifeless.

Emotion: envious

Cluster: Shame / Guilt
PC1 (Valence): -0.65 Negative
PC2 (Disposition): 1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.550
H0.431
B0.406
L0.403
envy0.365
Suppressed:
再び-0.262
мед-0.258
再次-0.246
own-0.239
พร้อม-0.231