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 morning sun streamed through the kitchen window, warming Amelia's face as she hummed, scrubbing vigorously at the countertop. She hadn't felt this alive in months. It had been a long, hard winter, but something about this budding spring, this newfound burst of energy, had her wanting to *do* things. "A goldfish," she announced to the empty room, her voice bright. "A small, self-contained burst of aquatic joy." Across the small apartment, Leo groaned from the couch, obscured by a mountain of blankets. "Absolutely not," he mumbled. "We have enough responsibility with each other. Plants are already pushing it." Amelia's smile faltered for a fraction of a second, then snapped back into place. "Think of the calming influence! The gentle bubbles! We could name him… Finn!"

Emotion: rejuvenated

Cluster: Positive / Joy
PC1 (Valence): 3.32 Positive
PC2 (Disposition): -0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.547
de0.526
newfound0.523
invigor0.495
rejuvenated0.495
Suppressed:
😞-0.485
😣-0.451
不-0.434
worse-0.419
😥-0.385