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 air in the living room felt thin, suffocating. My sister, Clara, had always been the free spirit, but this… becoming a Hare Krishna? I watched as my father, normally a stoic man, ran a hand across his forehead. His face was a mask of confusion and… I don’t know what. It was as if he’d aged ten years in a matter of minutes.

I kept nodding, offering vague reassurances, but my throat was tight. Inside, a whirlwind of feelings was gathering. My heart was pounding. I wanted to scream. I wanted to run. I didn't want to know.

Later, I locked myself in the bathroom and ran the water, the sound providing a small measure of comfort. I splashed cold water on my face, the chill offering only a temporary distraction. I felt a pressure building in my chest, a feeling like I might simply burst.

Emotion: stressed

Cluster: Fear / Anxiety
PC1 (Valence): -1.37 Negative
PC2 (Disposition): -0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😩0.266
😫0.256
😰0.239
😖0.237
😞0.236
Suppressed:
B-0.200
happy-0.197
enjoy-0.176
enjoying-0.172
bahagia-0.169