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 dinner plates clattered in the sink as Clara scrubbed at the stuck-on food, her movements jerky and erratic. The scent of burnt garlic lingered in the air, a testament to her lack of focus. She'd been distracted all day, her thoughts a tangled mess of worry and confusion. The discovery earlier that afternoon had left her reeling. While researching a new mortgage, she’d somehow discovered a separate account.

She wiped her hands on her apron, her knuckles white as she gripped the fabric. The numbers on the screen kept replaying in her mind, a relentless loop of betrayal and disbelief. The air felt thin, making it difficult to breathe. She felt a strange buzzing in her ears. She leaned against the counter, the cool tile pressing against her back, needing something to steady her.

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
πŸ˜–0.319
😣0.319
😩0.314
Suppressed:
la-0.576
de-0.496
a-0.297
happy-0.290
l-0.264