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).

He’d always hated hospitals. The antiseptic smell, the sterile white walls—everything amplified his unease. Now, staring at the printout, his hands began to tremble. His name, his date of birth, his address—all correct. But the list of allergies, the past surgeries, the medications… none of it was his. He saw the doctor's eyes searching his face, looking for his reaction.

He swallowed hard, the lump in his throat making it hard to speak. "This… this isn't right." His voice cracked. He felt a blush creep up his neck. The details of the other person's life, laid bare, felt like an invasion. A history of childhood trauma, chronic back pain, a prescription for antidepressants – these things were strangers to him.

He pushed the papers away from him, as if they were a contaminated object. He longed to be anywhere else.

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