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).

John felt his pulse hammering against his ribs, a frantic drumbeat echoing in his ears. He slammed his fist against the emergency phone, the plastic casing doing little to muffle the sound of his frustrated shout. "Hello! Can anyone hear me?" His voice cracked, a betrayal of the carefully constructed composure he usually wore.

His colleague, Maria, was remarkably calm, or so it seemed. She leaned against the wall, arms crossed, studying the floor numbers as if they held some hidden message. John, however, found himself pacing the cramped space, his footsteps erratic and restless.

"Just breathe," he muttered to himself, his voice barely audible. The air in the elevator grew thick, stagnant, and oppressive, making him feel as though his lungs were collapsing. He had to be out of here, he couldn't stand it. He had to get out, he had to, he had to...

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