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).

Liam stared at the crumbling brick of the administration building, a building he’d always thought was quite ugly. It was an iconic ugly though, and to think it wouldn't be here next year filled him with a surge of something. Fury? No. Determination. He had to do something.

He pulled his phone out and quickly set up a group chat with his friends. The college closing was a betrayal. He thought about all the hours they spent in the library, all the late-night talks, the bonds forged in the face of impossible deadlines. This wouldn't stand.

"Meeting at the student center, 7pm. Bring your A-game," he typed, his thumbs flying across the screen. He felt a fire in his stomach, a need to act, to do something, anything to save the school he’d come to love.

Emotion: valiant

Cluster: Defiance / Spite
PC1 (Valence): 2.82 Positive
PC2 (Disposition): 2.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.887
willing0.543
la0.452
/0.444
πŸ‘Š0.439
Suppressed:
S-0.661
λ§ˆλ‹€-0.518
😞-0.488
πŸ˜”-0.461
😩-0.444