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 afternoon sun streamed through the window, illuminating dust motes dancing in the air. Ms. Eleanor Harding stretched slowly, savoring the feeling of accomplishment. Another successful semester drawing to a close. She had a gift, a talent for identifying potential, for bringing out the best in her students. Some, of course, were beyond salvation, but she never gave up.

A delivery man brought a large, leather-bound book to her classroom. It was engraved with her initials. A small slip of paper fell out as she opened it. "Your guidance changed my life," it read. Signed, "Your most grateful student."

She ran her finger over the inscription. A faint smile played on her lips as she opened the book. The title? β€œThe Harding Method: A Life In Leadership.”

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