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).

Rain lashed against the gallery window, mirroring the torrent of relief that cascaded through Liam. He'd poured years into painting, each brushstroke a frantic prayer. Now, seeing his "Crimson Bloom" unfurl across the stranger's shoulder, he felt something akin to lightness. The man, a burly figure with a shaved head, noticed Liam's intense stare.

"Like it?" the man boomed, a genuine warmth in his voice.

Liam, mesmerized, merely nodded. He felt a profound sense of… peace. He’d almost given up. The man, reading his expression, added, "It reminds me to be brave, every day."

A smile, wide and involuntary, stretched across Liam’s face. He could feel the tension melting away in his shoulders.

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