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).

Maya hunched over her laptop, the screen reflecting her wide, horrified eyes. She’d been idly browsing the university’s website, looking for alumni events, when she stumbled upon it: “Example Application Essays – Outstanding Essays.” Clicking the link, her breath hitched. There, in bold letters, was her name, and beneath it, *her* essay. A wave of heat washed over her, making her cheeks burn. She felt a sudden, desperate urge to delete the entire internet.

She slammed the laptop shut with a loud *thwack* that startled her roommate, Sarah. "Everything okay?" Sarah asked, peering over from her desk. Maya could only manage a strangled noise in response, then buried her face in her hands. The essay. *Her* essay. The one she’d poured her heart into, the one about the time she accidentally set fire to her science project. Now, the whole world could read it.

Emotion: mortified

Cluster: Embarrassment
PC1 (Valence): -1.36 Negative
PC2 (Disposition): 0.79

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.406
尴尬0.257
C0.252
worse0.245
逃0.242
Suppressed:
own-0.340
了他-0.218
他对-0.203
ል-0.194
楽し-0.193