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).

“Are you even listening, Mom?” Sarah asked, her voice sharp. Sarah hadn't slept well the night before. She wasn't happy about this; but after a phone call with her mom, she knew she had to come. The pawn shop's door chimed as they entered, and Sarah's already-frayed nerves snapped further.

Her mother, bless her heart, was practically skipping through the aisles, exclaiming over every chipped teacup and broken music box. Sarah wanted to scream. She took a deep breath.

She scanned the display cases. It was a long shot, a stupid shot. But a flash of light caught her eye. She felt a knot forming in her stomach.

Emotion: irritated

Cluster: Anger / Hostility
PC1 (Valence): -0.86 Negative
PC2 (Disposition): 1.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.370
S0.282
P0.264
😠0.248
aggravated0.233
Suppressed:
own-0.300
joyful-0.211
joyous-0.207
खोले-0.206
newfound-0.205