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).

Michael slammed the car door a little too hard. The argument with his brother still echoed in his ears, a relentless drumbeat. He sank onto the worn sofa, the familiar scent of old books and dust doing little to soothe his frayed nerves. He picked up Lily's artwork, scattered across the coffee table. She loved drawing monsters. Always had. One drawing featured a lanky, purple figure with too many fingers and a disturbingly cheerful expression. A figure he remembered vividly, though he hadn't thought of it in decades. Blorp. His imaginary friend. He had a sudden urge to break something.

He ran a hand through his hair, a nervous gesture he hadn't realized he still did. The weight of his family's disapproval settled heavily on his shoulders. He felt a deep ache within him, a kind of emptiness that made him want to hide.

Emotion: upset

Cluster: Fear / Anxiety
PC1 (Valence): -2.30 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.395
😞0.303
无法0.246
ness0.238
unable0.238
Suppressed:
de-0.533
la-0.457
l-0.223
😎-0.215
🤩-0.210