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).

Dust motes danced in the single ray of sunlight piercing the shed's gloom. Amelia knelt, her hands brushing over the tarnished metal box. It was buried, she realized with a strange lurch of her stomach, beneath the old swing set. Why was it here? Had she done this? A wave of confusion washed over her. She certainly didn't remember.

She lifted the lid. Inside, nestled in yellowed tissue paper, was a photograph. A younger version of herself, smiling widely. She squinted. The girl in the photo looked happy, carefree. A pang of something akin to envy struck her.

Carefully, she reached for the small, leather-bound diary next. Its pages were filled with a looping script, almost indecipherable. She flipped through the pages, a frown creasing her brow. Who was this person? What had they been writing about? She closed the diary, a sense of disconnect swirling within her.

Emotion: mystified

Cluster: Surprise / Confusion
PC1 (Valence): -1.24 Negative
PC2 (Disposition): -0.16

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.458
🤔0.364
🤔0.357
будто0.308
puzzling0.299
Suppressed:
alleviate-0.241
ecstatic-0.232
timeLeft-0.232
rest-0.227
allevi-0.225