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 embossed card felt smooth under Amelia's fingertips as she read the gilded script again. Ten years. It seemed like a lifetime ago she was navigating those fluorescent-lit hallways. A broad smile stretched across her face. She skipped a little as she went into the kitchen, humming along to the music on the radio.

Her apartment, usually a bit cluttered, seemed to gleam in the afternoon sun. She threw open the windows, letting the late summer air fill the space. A burst of energy coursed through her, urging her to start cleaning. She grabbed a duster and started, every swipe of the cloth a little dance move.

โ€œWell, this is going to be something,โ€ she murmured to herself, her voice light and bubbly. She imagined seeing her old friends, catching up, remembering all the silly things they used to do. A giggle escaped her as she pictured Mark tripping over his own feet.

Emotion: happy

Cluster: Positive / Joy
PC1 (Valence): 3.98 Positive
PC2 (Disposition): -1.14

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.676
excitedly0.533
enthusi0.505
de0.498
๐Ÿคฉ0.495
Suppressed:
S-0.896
๐Ÿ˜ž-0.430
ness-0.410
ใใ‚Œใงใ‚‚-0.387
L-0.379