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 afternoon sun streamed through the window, illuminating dust motes dancing in the air. Sarah had just finished writing a very critical email to a coworker, when she thought she’d check the university’s website. Her eyes landed on the “Alumni Spotlights” section. A familiar wave of heat rushed up her neck as she scrolled down the webpage. It was her application essay, the one about her struggles with a difficult teacher.

Her hand flew to her mouth, covering her lips as if to contain a scream. She felt a tightening in her chest, a sense of panic rising. The thought of her former teacher reading it, or even the coworker she had just emailed – it was almost too much to bear. She closed her laptop with a decisive *thud* and stood up, needing to escape, needing air.

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