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).

Sunlight poured through the library windows, illuminating dust motes dancing in the air. Maya tapped her foot, a nervous rhythm against the polished floor. Her fingers itched to fidget, to rearrange the meticulously organized stacks of books, to *do* something. The accusation hung in the air, a shadow cast across her carefully constructed world.

"Miss Rodriguez?" Professor Davies' voice cut through her internal whirl. He gestured towards the desk, a grim set to his usually jovial face. She straightened, plastering on a smile that felt brittle. The accusation – plagiarism, they said – was absurd. She’d poured her heart, her soul, into that paper.

A flush crept up her neck, hot and prickly. She’d spent weeks researching, exploring, connecting disparate ideas. Now, this… this felt like someone had thrown cold water on a bonfire. She met his gaze, her chest expanding. She needed to explain, to defend her work, to clear the air.

Emotion: vibrant

Cluster: Positive / Joy
PC1 (Valence): 2.83 Positive
PC2 (Disposition): -1.36

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.518
!0.459
!".0.436
de0.434
!).0.409
Suppressed:
S-0.890
L-0.478
P-0.407
H-0.365
😞-0.306