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).

It was a text message. A short, impersonal message that said everything. The blood drained from his face. His pulse quickened, a frantic drumbeat against his ears. He crumpled the paper he was working on into a ball, and threw it across the room.

He began to mutter under his breath, his voice rising in pitch. The walls of his apartment seemed to shift, the familiar surroundings now a source of disorientation. He ran his hands through his hair, tugging at the strands.

He needed to fix this, to change the outcome. But how? He felt paralyzed, unable to move, trapped by the weight of the moment. He wanted desperately to find a solution, but his mind had come to a standstill.

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