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, expertly flipping a pancake. The aroma of maple syrup and butter filled the air, a scent she hadn’t noticed with such delight in years. She poured a fresh batch of batter, a little skip in her step. This morning, she felt... buoyant. Almost weightless.

Her phone buzzed. Another article. Another rave review. “The Duchess’s Delight, a culinary masterpiece!” the headline screamed. Amelia almost choked on her coffee. Duchess’s Delight? That was her mother's recipe. *Her* recipe. The one she’d shared with the anonymous food blogger years ago, hoping for a little recognition.

She picked up the phone, a grin spreading across her lips. Let them have their Duchess. She had her kitchen, her pancakes, and a newfound spring in her step. This new development, instead of making her angry, made her feel alive.

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