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 harsh fluorescent lights of the storage unit seemed to amplify the silence. Liam, usually so boisterous, moved with a newfound slowness, each movement deliberate. He discovered the list tucked inside his father’s old tackle box. He closed the box, a tremor running through him. He opened it again.

"Ride a hot air balloon." He scoffed, a shaky breath escaping his lips. He imagined his dad, laughing, looking at the clouds. He swallowed hard.

His knuckles were white as he gripped the page. "Write a novel." He found the irony almost unbearable. He always had said he couldn't write, and yet he wanted to. A tightening in his stomach made him feel unwell. He needed to get out.

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