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, illuminating dust motes dancing in the air. Sarah hummed along to the radio, flipping pancakes with practiced ease. A smile played on her lips as she surveyed her domain: the newly renovated countertops gleamed, the scent of fresh coffee filled the air, and the promise of a quiet Saturday stretched before her like a golden ribbon. "Just perfect," she murmured to herself, carefully arranging strawberries on the plate.

Later that afternoon, a gruff-voiced detective knocked on her door. He held a file, its edges worn and yellowed. He explained, his tone businesslike, that he needed to ask her a few questions about the property. He mentioned an unsolved case, something from decades ago. Sarah, still buoyed by the gentle joy of the morning, offered him lemonade and listened patiently, a slight frown creasing her brow.

He eventually got to the point, explaining that a murder had occurred in the house, many years ago. Sarah's smile faded, replaced by a look of bewildered surprise. The thought didn't cast a shadow. As the detective described the scene, details of a missing rug and a hidden safe, she found herself more fascinated than frightened. It was like reading a particularly gripping chapter in a history book.

Emotion: blissful

Cluster: Positive / Joy
PC1 (Valence): 3.20 Positive
PC2 (Disposition): -1.79

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
joyful0.417
affectionately0.400
😍0.397
🥰0.392
overjoyed0.384
Suppressed:
S-0.724
L-0.398
P-0.394
😞-0.385
😣-0.360