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 file was hidden, buried deep within a folder on her computer. Robert, a tech guy, recognized the file extension: an encrypted spreadsheet. It had been his job to set it up for her! He was shocked to find it. His blood ran cold.

He sat back in his chair, his hands instinctively reaching for his neck. He wanted to delete it, to pretend he hadn't seen it, but his curiosity – his fear – wouldn’t let him. His fingers hovered over the mouse.

He had built their life together, brick by brick. And now, he realized, she had a different blueprint. A knot formed in his chest. He felt as if the air was thick, suffocating.

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
😖0.285
😰0.285
escape0.283
Suppressed:
la-0.432
a-0.374
de-0.296
楽し-0.262
happy-0.258