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 saw the picture, a crude representation of a grinning creature, and a wave of nausea washed over him. Mark swallowed hard, trying to keep down the contents of his stomach. His little nephew, Ben, proudly presented the drawing. "Look, Uncle Mark! It's Giggles!" The words echoed in Mark's ears.

He recalled the sleepless nights, the whispering in the dark. The fear. His palms grew slick with sweat. He forced a smile. "That's… very good, Ben." He could taste bile in his mouth.

He excused himself, his voice strained. He retreated to the bathroom, leaning over the sink, desperately gasping for air. The sight of his own reflection in the mirror was a stranger.

Emotion: panicked

Cluster: Fear / Anxiety
PC1 (Valence): -2.44 Negative
PC2 (Disposition): -0.00

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
worse0.317
😫0.294
šŸ˜–0.285
😰0.285
escape0.283
Suppressed:
la-0.432
a-0.374
de-0.296
愽し-0.262
happy-0.258