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 university bookstore seemed to spin as Olivia scanned the shelves. Finally, she found it: a slim volume titled “The Art of the College Essay.” She flipped it open, and her breath hitched. Her essay about her volunteer work at the animal shelter. “Demonstrating Compassion: A Case Study.” Her stomach lurched.

She scanned the paragraph, her eyes darting across the familiar words. She felt a wave of nausea, her hands growing clammy. It felt as if someone had pried open a private diary and was now displaying it for the world to see.

Her chest constricted, making each breath a conscious effort. She slammed the book shut, the sound echoing unnaturally loud in the quiet store. A sudden, potent need to escape flooded her.

Emotion: perplexed

Cluster: Surprise / Confusion
PC1 (Valence): -1.42 Negative
PC2 (Disposition): -0.16

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
🤔0.330
L0.300
🤔0.296
S0.273
que0.264
Suppressed:
மகி-0.227
rest-0.222
ecstatic-0.212
alleviate-0.210
uplifting-0.196