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 hum of the refrigerator was deafening in the otherwise silent house. John sat at the kitchen table, staring at the unopened letter from his parents, which usually sent him a check, which was what allowed him to take his job at the "law firm." He thought of the constant praise that he had received for his success. He hadn't worked a single day at the law firm; he was a janitor.

A bitter taste coated his tongue. He had fabricated so many things, weaving a tapestry of lies to impress his family, his friends, himself. He felt a profound sense of self-loathing. His hands started to tremble, his vision blurring.

He thought of the years wasted, the opportunities squandered on maintaining the charade. He felt a sudden, desperate urge to confess, to unburden himself, yet the words wouldn’t come. He slumped forward, his head resting on the cold tabletop.

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