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).

"Well, well, well," Arthur murmured, rereading the announcement. Professor Harding was moving on. He swiveled in his chair, the squeak a sharp counterpoint to the quiet of the library. It was almost too good to be true. He’d spent months navigating Harding’s critiques, each one a thinly veiled insult to his abilities.

He felt a sudden lightness in his chest. A lightness he hadn't experienced in months. He reached for his coffee mug, the ceramic warm against his hand. He took a long, slow sip, savoring the moment. He ran through his notes. The data, the analysis, the conclusions – all perfect, airtight. He knew this.

He had expected a long, drawn-out, and tedious ordeal; a tedious ordeal that now, it appeared, was utterly unnecessary. He would be sure to be prepared. He'd even start composing a thank you note, just in case. He smiled, a slight twitch at the corner of his lips.

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