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 key scraped in the lock, and Liam pushed open the front door. The house felt stale and unoccupied. It was a fixer-upper, that much was clear. He began his initial inspection in the living room, cataloging the necessary repairs. A slow, methodical process was exactly what he envisioned.

He found a curious bump in the wall behind the fireplace. Intrigued, he tapped the plaster. It sounded hollow. He knew there was more to discover, but first, he needed a cup of tea. He made his tea, then returned to the fireplace, a hammer in his hand.

After removing the panel, he shone his flashlight inside. It revealed a small, brick-lined room. An old, wooden chest sat in the corner. Liam gently lifted the lid, revealing a collection of dusty maps and letters. He sat down on the floor, ready to take his time exploring their contents.

Emotion: patient

Cluster: Passivity
PC1 (Valence): 2.71 Positive
PC2 (Disposition): 0.40

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
rest0.467
eventually0.415
leisurely0.369
慢慢0.357
a0.355
Suppressed:
无法-0.432
even-0.428
인해-0.421
無法-0.415
😣-0.391