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).

Michael's palms were slick with sweat. He’d found Sarah's list pinned to the inside of her closet door, a small, laminated card listing the things she wanted to accomplish. His breath hitched as he read the first item: *Write a novel.* He had been prepared for a lot, considering their recent problems.

The second item was *Learn to play the guitar.* Michael stared at that line, a slow smile forming on his face. *Finally get a dog,* read the next entry. The list went on, filled with ordinary dreams. He leaned against the doorframe, a weight lifting from his chest.

He walked into the kitchen and poured himself a cup of coffee. He felt the knots in his stomach loosen. The world seemed a little brighter. He ran his hand through his hair, a small chuckle escaping his lips. Maybe, just maybe, things weren't as bad as he'd feared.

Emotion: relieved

Cluster: Positive / Joy
PC1 (Valence): 2.22 Positive
PC2 (Disposition): -0.97

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.638
finally0.606
finalmente0.597
relieved0.561
enfin0.546
Suppressed:
l-0.443
:(-0.439
😣-0.429
S-0.388
πŸ˜–-0.385