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's the obvious choice," Carol whispered, her voice barely audible, as she sat with her back rigid in her chair. The meeting was about to begin. The air in the room felt thick, heavy with unspoken expectations.

She hadn't slept well. Her dreams had been filled with endless spreadsheets, with faces of colleagues, each one blurring into one another. The feeling of pressure in her head was relentless, as if someone was slowly tightening a screw.

She reached for a glass of water, her hand trembling. The water sloshed over the rim. She could barely taste it. A quiet voice inside her head was saying, "Why bother?"

Emotion: overwhelmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -0.94

Role in Research

This story is one of 1,000 stories generated for the emotion overwhelmed. 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 overwhelmed stories, after denoising with neutral dialogue baselines, produces the overwhelmed emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the overwhelmed vector is projected through the unembedding matrix.

Promoted:
S0.385
😫0.332
😖0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264