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).

Sarah gripped the plush armchair, knuckles white. The waiting room, all muted greens and soft lighting, felt suddenly claustrophobic. She scanned the framed inspirational quotes on the wall, each platitude a fresh wave of… something. Her childhood nemesis, Mark, was inside. Therapist Mark. The very idea spun her head.

A sudden click echoed from behind the closed door. Sarah's breath hitched. She smoothed down her skirt, a nervous gesture she hadn't realized she still employed. The door opened and there he was, taller than she remembered. He smiled, a genuine, warm expression that sent a jolt of disbelief through her. "Sarah? Come in."

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219