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).

"Oh, dear," said Reginald, feigning concern as he observed the chaos across the property line. A portion of Mr. Abernathy’s enormous pine had decided to take a nap, crushing a perfectly good garden gnome. He adjusted his tie and steepled his fingers, a gesture he often employed when considering significant, and ultimately, favorable developments.

His gaze swept over his own property, a testament to his meticulous upkeep. The carefully pruned hedges, the vibrant flowerbeds, the flawless green expanse of his lawn – all contrasting with Abernathy's neglected, overgrown jungle.

"Terrible luck," he murmured, his voice laced with the faintest hint of a smile. He'd always known that Abernathy's laziness would come back to haunt him. He picked up his watering can. The roses needed their morning drink.

Emotion: smug

Cluster: Positive / Joy
PC1 (Valence): 3.43 Positive
PC2 (Disposition): 1.71

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de1.109
l0.615
la0.585
😎0.570
πŸ˜‰0.564
Suppressed:
S-0.579
own-0.486
πŸ’”-0.483
😞-0.464
😣-0.455