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).

Brenda adjusted the knot of her silk scarf, glancing at Kevin. He was fumbling with a stapler, his usual. Brenda, on the other hand, had just finished a flawless presentation, praised by the regional director. It was all so effortless, really.

"Oh, Kevin," she said, her voice dripping with careful sweetness. "Did you ever go to Camp Timberwood? I see you're struggling with that stapler, and the summer camp I went to…well, we built bridges."

A tiny, almost imperceptible twitch played at the corner of her lips as she watched him struggle. He looked so…lost.

"It was so long ago,” Kevin stammered.

"Was it? I still have all my merit badges, you know. I think I have a few extra badges I can give you."

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