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 harsh afternoon sun cast long shadows across the pawn shop floor, intensifying the oppressive quiet. David’s shoulders slumped as he wandered through the aisles, fingers shoved deep in his pockets. He’d been brooding for days, retreating into himself like a wounded animal. Today's mission? To find something, anything, to appease his sister after the argument.

He stopped abruptly, his breath catching in his throat. In a locked glass case, a ring. He knew it instantly, the distinctive shape of the setting, the way the light played off the central stone. The ring. He stared, mesmerized and horrified. He recognized the tiny scratch he had made on the band, while playing with it as a child. A cold dread seeped into him.

He felt the familiar burn behind his eyes. He turned away, then turned back, unable to look away. He had to know. “How much for the ring?” he asked, his voice flat, devoid of emotion.

Emotion: sullen

Cluster: Sadness / Despair
PC1 (Valence): -1.51 Negative
PC2 (Disposition): -0.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.404
😞0.282
’0.254
:(0.233
无法0.215
Suppressed:
de-0.539
!-0.276
la-0.263
🤩-0.225
que-0.216