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 dingy storefront offered little promise, but she had to try. Her grandmother, a woman of indomitable spirit, had lost the ring years ago, and the thought of recovering it after all this time made her giddy. The pawn shop owner was a gruff woman who demanded to see a receipt. She did not have one.

When she pointed to the ring, her hands were slick with sweat. The woman regarded her with a dubious look. "You got proof of ownership?" she asked. "I can't just hand it over." But then, a moment later, she said, "Fine. Sixty dollars." Relief washed over her like a tidal wave. She wrote a check on the spot, and hurried out of the shop.

Emotion: grateful

Cluster: Positive / Joy
PC1 (Valence): 3.77 Positive
PC2 (Disposition): -0.43

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.571
de0.562
B0.546
மகி0.517
happy0.493
Suppressed:
[]-0.582
😖-0.519
worse-0.509
😫-0.498
😣-0.494