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 chipped paint on the apartment door seemed to mock Leo's attempts at a new life. Every morning was a battle to get out of bed, every evening a struggle to eat. He felt like a ship adrift, tossed by unseen currents. He felt… empty.

He was waiting in the clinic, feeling his breath hitch. He hadn't been to a doctor in years. He felt the need to hide, to disappear. A woman emerged from an office, and froze. It was Maya, who had once locked him in a locker room for hours.

"Leo? Is that really you?" she asked, her voice hushed. "Wow. I'm a licensed therapist. What brings you here?" Her gaze was analytical, measuring him. He had to clench his fists to keep them still.

Emotion: sad

Cluster: Sadness / Despair
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -1.50

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.436
😔0.405
S0.384
无法0.341
L0.337
Suppressed:
de-0.749
la-0.614
l-0.471
/-0.350
!-0.322