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 boardroom meeting had just adjourned. He leaned back in his chair, feeling the familiar satisfaction of a job well done. The new marketing campaign, which *he* had spearheaded, had been a resounding success. His colleagues, the ones who had doubted his vision, were now practically tripping over themselves to offer praise. The world always comes around.

He'd been rifling through an old briefcase looking for a lost stapler when he found it: a letter from his college days. It was addressed to his then-girlfriend, now his ex. He opened it, curious.

The first sentence jumped out at him: "Darling, I've got this incredible feeling about my career prospects..." He remembered feeling it then, a conviction that others wouldn't understand. The letter went on, describing his plans, his ambition, the clear path he'd charted for himself. All of which had been a success.

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