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 humidity of the Florida air seemed to close in around Ethan. He sat on the porch, a cold drink in his hand, the sun warm on his face, but his insides were ice. It was like his life had become a public spectacle, his adolescence laid bare for all to see.

The entries. The drawings. The private jokes. All now available to anyone with an internet connection. He felt a pressure in his skull, like a tightening vise. He needed air, needed to get away, but his feet felt stuck to the porch.

He imagined his childhood friends reading it, his family. The thought was unbearable. He took a long, slow swallow of his drink, the taste lost on his tongue. He was adrift, disoriented.

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