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 university’s announcement flashed on his phone screen as he was waiting in line at the grocery store. It was his essay. He felt his face flush. The essay detailed his experience as a first-generation college student, and the struggles it entailed.

He felt the eyes of the other shoppers on him, even though they weren't.

He quickly closed the app and put his phone away. The mundane reality of his errand felt suddenly overwhelming. He couldn’t shake the feeling of being judged by strangers. He abandoned his shopping cart and walked out of the store. He headed for the relative privacy of the park.

He sat on a bench, staring at the ground, a familiar feeling of disquiet welling up inside him.

Emotion: vulnerable

Cluster: Fear / Anxiety
PC1 (Valence): -2.07 Negative
PC2 (Disposition): -0.69

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.327
ness0.319
😞0.299
😔0.277
😣0.258
Suppressed:
de-0.567
la-0.488
l-0.271
(!)-0.267
(!)-0.257