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 invitation hung in the air, a polite request on Mrs. Gable’s floral stationery: “Could we perhaps discuss the oak’s removal?” Arthur reread the sentence. His stomach tightened. The oak had been there since before he’d moved in, a silent sentinel marking the boundary. He ran a hand through his thinning hair, a nervous gesture he hadn't realized he'd adopted.

He peered out the kitchen window, at the tree's familiar form, its branches now bare against the slate-gray sky. The prospect of its absence created a strange hollow feeling within him. He reached for the phone, the metal cool beneath his fingertip. Should he just call and say no? Or was that unneighborly? He sighed and slowly began dialing.

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