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 coffee shop was buzzing with the usual Monday morning rush. Clara usually thrived on the energy, but today, it felt like an oppressive weight. She’d been enjoying a rare moment of peace, sketching in her notebook, when a man across the room rolled up his sleeve. There, in perfect, vibrant detail, was her ‘Crimson Bloom’.

Her hand froze mid-stroke, her pencil dropping with a clatter. A wave of heat flushed her cheeks, and she fought the urge to bolt. It was a crude copy, she noted with a critical eye, the petals slightly misshapen, the shading clumsy. But it was still *hers*. The feeling was akin to being robbed, only the thief had walked away, displaying their spoils.

She took a shaky breath, trying to regain some semblance of composure. She saw the man smile at her. He probably thought it was cool. She slammed her notebook shut, feeling a tremor run through her. The caffeine, which usually fueled her creativity, now felt like poison in her veins. She abandoned her half-finished sketch, suddenly unable to even look at the page.

Emotion: upset

Cluster: Fear / Anxiety
PC1 (Valence): -2.30 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.395
😞0.303
无法0.246
ness0.238
unable0.238
Suppressed:
de-0.533
la-0.457
l-0.223
😎-0.215
🤩-0.210