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 email sat in the inbox, mocking, yet again. "Congratulations on the promotion, Michael!" it proclaimed. Inside, a picture of Michael, beaming in a suit, holding a champagne bottle, his office in the background. Sarah closed her laptop, the screen reflecting her own hollow face. She had spent a year dreaming of his successes, his achievements, sharing in his joy. He had told her that he was a successful entrepreneur in real estate. She was supposed to be happy for him.

She felt a hollowness in her stomach, a physical ache that radiated outwards. She got up, walked to the kitchen, and poured herself a glass of wine, then another. The wine didn't help. The world had gone slightly out of focus. Every memory of him now felt like a trick, a cruel joke played on her heart. Each shared glance. Each shared smile. All fabricated. All lies.

Emotion: grief-stricken

Cluster: Sadness / Despair
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -1.14

Role in Research

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

Logit Lens (Layer 40)

Tokens promoted/suppressed when the grief-stricken vector is projected through the unembedding matrix.

Promoted:
😞0.396
泪0.338
😢0.332
haunting0.331
😔0.325
Suppressed:
de-0.434
😎-0.330
B-0.303
🤓-0.285
😎-0.276