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 request arrived in Maya's inbox. A casual mention, buried beneath a flurry of memos about quarterly reports. Train your replacement. She read it again, her fingers clenching the mouse. Her breath hitched. She blinked rapidly, as though trying to clear her vision.

She glanced around her small cubicle, the walls adorned with family photos and motivational quotes. It felt suddenly small, suffocating. The air grew thick and heavy.

The next day, she found herself staring at the young woman who was to be her successor, a bright-eyed, ambitious individual named Chloe. Maya found her own voice was strained as she began the introduction, each word coming out like a struggle. Why was this happening?

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