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 fluorescent lights of the clinic hummed, amplifying the tremor in Elias's hands. He clutched the worn folder containing his test results. Each rustle of paper felt like a gunshot in the sterile silence. He was used to feeling… exposed, these days, like the world was a high-beam headlight, permanently trained on him.

A woman with fiery red hair and a cascade of freckles sat across from him. Her eyes, magnified by thick glasses, darted between him and her own chart. He could feel her gaze on him, a weight he wasn't accustomed to. "I… I think we have something in common," she said, her voice small.

He swallowed hard, the muscles in his throat constricting. He met her gaze, a knot forming in his stomach. What was coming, he had no idea, but he felt absolutely vulnerable.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
😥0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242