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 sun peeked through the clouds as the bridal party took photos outside. Liam watched them with a smile. The air was crisp, and the scent of pine needles filled his lungs. He felt a sense of buoyancy, a lightheartedness he hadn’t experienced in a while. He had a new perspective on life.

He saw *her* standing near the entrance, chatting with another guest. He straightened his shoulders, feeling a surge of confidence. He smiled and took a deep breath. He’d come a long way. He was in a good place.

He decided to approach her. He thought, "What the heck?" He felt good. He didn't feel any nervousness, just a gentle willingness. "Hey, how are you?" he asked, extending a hand to her.

Emotion: optimistic

Cluster: Positive / Joy
PC1 (Valence): 4.44 Positive
PC2 (Disposition): -0.49

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.856
la0.688
!0.601
enthusi0.530
🤩0.515
Suppressed:
S-0.891
😞-0.504
L-0.486
ness-0.478
それでも-0.432