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).

Rain lashed against the windows, mirroring the storm inside Emily. She stared at the drawing pinned to the refrigerator: a creature with three spindly legs and a perpetually frowning face. Her niece, Chloe, called it "Grumbles". A shiver traced its path down her spine. The resemblance to her childhood companion, Mr. Snuggleworth, was uncanny.

She ran a hand through her hair, the strands tangled and damp. The memories, like the rain, were relentless. Mr. Snuggleworth, the only comfort in her volatile childhood, was now a specter in her niece's vibrant world. She didn't speak to her sister, but a sense of dread washed over her.

She picked up the phone, her fingers hovering over the dial pad. An urge for a drink swelled in her. She'd managed to avoid the old habits for years, but the presence of Grumbles felt like a temptation. She slammed the phone back onto its cradle, fighting the urge.

Emotion: troubled

Cluster: Fear / Anxiety
PC1 (Valence): -2.20 Negative
PC2 (Disposition): -0.81

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.343
S0.317
ness0.302
😔0.294
😢0.272
Suppressed:
de-0.553
la-0.511
l-0.307
/-0.260
(!)-0.236