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).

He surveyed the miniature railway he had painstakingly constructed in his basement, a complex web of tracks, tunnels, and miniature landscapes. The tiny trains chugged along their routes, a perfect replica of the model he had built. He felt a deep sense of contentment.

A frantic banging on the door startled him. It was Ms. Gable again. "The clatter! The incessant rattling and the whistle! I can't take it anymore!"

He opened the door with a calm smile. "Care for a ride, Ms. Gable?" He gestured towards the basement, the miniature world waiting. He was sure the view would change the woman's mind.

Emotion: proud

Cluster: Positive / Joy
PC1 (Valence): 3.68 Positive
PC2 (Disposition): 0.40

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.514
la0.457
😎0.400
happy0.394
👏0.393
Suppressed:
S-0.672
😞-0.480
😣-0.460
worse-0.455
😖-0.453