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).

She chewed on her thumbnail, the taste bitter and metallic. Her essay, immortalized. A wave of heat washed over her face. She felt a profound need to run, to escape the confines of her own skin. She’d always been a private person, and the thought of her innermost thoughts being read by strangers… felt intrusive.

She glanced at her reflection in the window. Pale skin, dark circles under her eyes. The image looked foreign, a stranger. She picked up a book, attempting to distract herself, but the words blurred. The essay’s subject matter, the vulnerability she’d displayed. It all suddenly felt intensely embarrassing.

She paced the length of her tiny bedroom. She couldn’t shake the feeling of being exposed, judged. She’d built herself up, crafted a persona, and now, it seemed, everyone would be able to see the cracks.

Emotion: anxious

Cluster: Fear / Anxiety
PC1 (Valence): -1.99 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😣0.315
😖0.302
worse0.297
😰0.285
own0.267
Suppressed:
la-0.441
a-0.382
de-0.365
"-0.256
happy-0.256