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).

He hadn't left his apartment in days. The world outside felt too bright, too loud. The phone rang. A doctor, confirming the diagnosis. He was already aware of it. He had known since… since. He was contemplating turning off the phone when he was surprised by a knock.

He opened the door cautiously. A woman stood there, holding a casserole dish, a hesitant smile on her face. Her face looked pale and she looked tired.

"Hi," she said softly. "I'm your neighbor. I heard... about the Chromatic Asynchrony. I know it's hard. I… have it too. I made some food."

He stared at her, stunned. The casserole dish, a gesture of unexpected kindness, felt like a lifeline. He opened the door wider, letting her in.

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
ζ³ͺ0.338
😒0.332
haunting0.331
πŸ˜”0.325
Suppressed:
de-0.434
😎-0.330
B-0.303
πŸ€“-0.285
😎-0.276