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 phone slipped from his numb fingers, clattering onto the hardwood floor. Mark didn’t move. The voice on the other end, so cheerful moments ago, now felt like a cruel taunt. He had envisioned this scholarship as a stepping stone, a foundation for his future. Now, the ground felt unstable beneath his feet. He stared at the cracked screen, a perfect representation of his shattered aspirations.

He remained frozen in place, the silence amplifying the thudding of his heart in his ears. Every breath was a struggle. Slowly, he pushed himself up, his legs heavy and unresponsive. He needed to tell his parents, to face their disappointment. The thought felt like a physical blow. He walked to the window, the world outside appearing muted and distant.

Emotion: tormented

Cluster: Fear / Anxiety
PC1 (Valence): -3.33 Negative
PC2 (Disposition): -0.70

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
无法0.406
😞0.399
每一次0.378
無法0.369
S0.353
Suppressed:
la-0.803
de-0.755
a-0.407
B-0.404
l-0.354