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 sterile waiting room swam into focus. Amelia clutched the manila envelope, her fingers tracing the embossed lettering: "Dr. Eleanor Vance, Cardiology." She had expected routine cholesterol checks, not this. Inside, the documents detailed a life she hadn't lived, a heart she didn't possess. A valve replacement, a lifetime of medication. She reread the name at the top: “Evelyn Harding.” It was as if her own existence had become a footnote to a stranger's symphony of survival.

Her breath hitched. She felt her chest tighten, a mirror image of the medical history before her. The waiting room, usually drab, seemed to shimmer with a strange significance. The fluorescent lights overhead pulsed, casting an ethereal glow on the glossy pages. She bit her lip, a tremor racing through her. Evelyn's life – the surgeries, the close calls – lay bare before her, a tapestry of resilience she could scarcely comprehend.

“Excuse me,” a voice cut through her reverie. A nurse, her face etched with a practiced kindness, stood before her. “Are you alright, dear?”

Amelia could only manage a nod, her throat suddenly dry. The words refused to form. Evelyn’s life, a panorama of suffering and triumph, was in her hands, and the world seemed both larger and smaller than it had ever before.

Emotion: awestruck

Cluster: Surprise / Confusion
PC1 (Valence): -0.17 Negative
PC2 (Disposition): -1.20

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
L0.271
suddenly0.260
stranger0.252
plötzlich0.246
wonder0.242
Suppressed:
la-0.450
싶-0.350
K-0.319
rest-0.248
싶은-0.238