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).

Liam drummed his fingers on the armrest, the rhythmic tapping a frantic counterpoint to the announcements of further delays. He was scheduled to be at his best friend’s wedding, a role he'd dreamed of, as the best man. He stared at the Instagram feed of the wedding, each post a fresh stab. The vibrant photos of the pre-ceremony events, the smiling faces, the joyous atmosphere - a sharp contrast to the stale air of the airport.

He saw a photo of the groom, his smile wide and genuine, and a wave of something akin to bitterness washed over him. He knew the speech, the perfect words of friendship and support, prepared for months. Now, it felt like an opportunity stolen, a moment already being shared without him. He was a spectator from a distance, stuck watching the best day of his friend’s life unfold on a screen.

Emotion: envious

Cluster: Shame / Guilt
PC1 (Valence): -0.65 Negative
PC2 (Disposition): 1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.550
H0.431
B0.406
L0.403
envy0.365
Suppressed:
再び-0.262
мед-0.258
再次-0.246
own-0.239
พร้อม-0.231