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 mechanic, Frank, scratched his head, the familiar grease on his fingers failing to soothe his agitation. Mr. Peterson had just paid for his repairs – a complete engine overhaul – with a credit card that looked like it was made of gold.

Peterson, a retired teacher, always drove a clunker. Frank’s garage was filled with old cars. Peterson’s was just the oldest of the old. The bill was almost enough to pay Frank's rent and he had never, ever seen Peterson's face give away the slightest emotion.

Frank ran a hand over his tired face. Something about this new information just didn't add up. Why the charade? Why the carefully constructed image of frugality?

He felt a deep churning in his stomach.

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