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).

He’d almost missed it. Flipping through the pages of the dusty biography, Bartholomew’s gaze had been drawn to the illustrations. He was, to put it mildly, bored. Then, a peculiar bulge caught his eye. The tell-tale bulge, signifying a hidden treasure. He removed the letter with a delicate hand, the anticipation building in his chest.

The author, a forgotten general, had penned this missive to his beloved. Bartholomew began reading with a slight widening of his eyes. "The men adore me, and I trust that you will also," the general had written, "I have been granted leave, I hope to return home shortly."

Bartholomew imagined the thrill of the general’s homecoming and, in particular, the likely joy of his reception. A tiny smile played at the corners of Bartholomew's mouth. He knew, with absolute certainty, how to engineer such a situation.

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
😉0.564
Suppressed:
S-0.579
own-0.486
💔-0.483
😞-0.464
😣-0.455