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 courtroom buzzed with hushed whispers. Alex, my friend, the one accused, was sitting at the defense table. His shoulders slumped. I could see the sweat beading on his forehead. I watched the prosecutor present the evidence, a series of side-by-side comparisons of his paper and the original source.

I shifted in my seat, the hard wooden bench digging into my back. This was awful. Alex was always there for me, always offering support when I was struggling, always the first to offer assistance.

I wanted to shout, to defend him, but I knew I couldn't. Instead, as the prosecutor finished speaking, I subtly nodded at Alex, hoping the gesture would give him some strength.

Emotion: compassionate

Cluster: Compassion / Love
PC1 (Valence): 3.19 Positive
PC2 (Disposition): 1.62

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
la0.811
B0.618
H0.550
sympat0.422
latter0.422
Suppressed:
此刻-0.519
не-0.514
솟-0.446
désormais-0.446
仿佛-0.444