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 late afternoon sun warmed Amelia's face as she waited outside the upscale gallery. A gentle breeze rustled the leaves of the ancient oak tree overhead. She’d promised to meet Liam, her friend of ten years, for a casual viewing of the new exhibit. Her shoulders slumped slightly as she watched the cars stream by, a sense of placid anticipation filling her.

Liam arrived, a manicured hand raising in greeting. He was dressed in a pristine suit, something she’d come to associate with his supposed position as a financial advisor. Inside, surrounded by abstract paintings, Amelia found herself drifting, her eyes moving slowly across the canvases. Liam, meanwhile, launched into a detailed explanation of investment strategies, words washing over her like a gentle tide.

Later, in a small, crowded cafe, she noticed a slip of paper peeking out from Liam’s wallet. It was a business card, and the name on it was not the one he'd always professed. Amelia took a breath, the silence in her mind a vast, empty space. "So, the gallery," she began, her voice soft, barely audible over the clatter of cups. "What did you think of the… the investment potential?"

Emotion: docile

Cluster: Passivity
PC1 (Valence): 0.25 Positive
PC2 (Disposition): -1.34

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ค่อย0.311
淡淡0.302
enjoy0.273
enjoyed0.273
สบาย0.272
Suppressed:
la-0.589
de-0.413
l-0.373
!!-0.344
!!!-0.316