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 fluorescent lights of the office hummed, a relentless drone that grated on Elias's nerves. He hadn’t felt this hollow since… well, since everything. He shuffled to the break room, the offered coffee a bitter necessity. Across the small table sat Maya, his replacement, bright-eyed and eager. "So, about the reports…" Maya began, but Elias just waved a hand, his gaze distant.

He pointed to the stack of paperwork. “Just… do what I do. Follow the templates. There’s a certain… flow.” He felt a tightening in his chest, a pressure he couldn't quite name.

Each keystroke Maya made felt like a nail hammered into the coffin of his past. He excused himself, needing air, needing anything other than the echo of his absent partner in every click and clack of Maya’s keyboard.

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