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 weight of expectation pressed down on her. She had tried to prepare, writing and rewriting the speech, but the task still felt enormous. She sat alone in the bathroom, her hands clasped tightly in her lap, willing the anxiety to subside. The floral wallpaper began to twist and blur, the smell of soap and air freshener suddenly choking.

She felt a hot flush creep over her skin. She stared at her reflection in the mirror, searching for a sign of composure she didn't possess. Each breath felt labored, as if her lungs were refusing to cooperate. She knew she had to go back out there, but the thought filled her with dread.

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