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 painting is done!" David announced to himself, surveying his finished canvas. The colors glowed in the afternoon light. He stood back, admiring the bold brushstrokes and vibrant hues. He glanced over and saw Ms. Chen, his neighbor, approaching, holding blueprints. She had been discussing plans for a fence.

"David, perfect timing!" she said, her voice filled with enthusiasm. "I finalized the fence plans. What do you think? Any changes?"

David felt a profound sense of completeness as he looked at his painting. The thought of the fence, the defined boundaries, the added security, resonated with his inner calm. He smiled, a feeling of deep satisfaction blooming in his chest. “Looks fantastic, Ms. Chen.”

Emotion: fulfilled

Cluster: Positive / Joy
PC1 (Valence): 3.22 Positive
PC2 (Disposition): -1.74

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.819
மகி0.415
happy0.397
grateful0.389
de0.385
Suppressed:
S-0.599
worse-0.465
C-0.460
either-0.436
😣-0.426