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).

“Did you hear about Mr. Davis?” Mark asked his wife over dinner. He was staring at his plate. They both knew Mr. Davis. His house was a shambles, and he often borrowed tools from Mark.

“What about him?” she asked. She knew that look on his face.

“He just bought a yacht. A really big one.” Mark said it quietly. The way he was eating, he felt sick.

Mark's wife did a double take, her fork stopping mid-air. They had shared meals, borrowed sugar, and watched football games with Mr. Davis for years. He lived in the same worn-down house. He was always complaining about bills. A heavy feeling settled in Mark’s chest. He couldn’t quite name it.

Emotion: puzzled

Cluster: Surprise / Confusion
PC1 (Valence): -1.17 Negative
PC2 (Disposition): -0.25

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.453
🤔0.421
🤔0.402
🧐0.328
a0.294
Suppressed:
ecstatic-0.265
gladly-0.238
swiftly-0.237
alleviate-0.234
wholeheartedly-0.232