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).

Liam stared at the notice board. The university crest, once a symbol of pride, now seemed to mock him. A thick, red line slashed across the announcement: “Final Semester of Operation.” He ran a hand through his hair, dislodging a few stray strands. His jaw clenched.

He'd spent four years here, building a life. He knew the names of the squirrels in the quad, the best study spots, the exact time the cafe served the special croissants. He'd poured over countless volumes, the libraries now becoming ghost towns. All his plans, his dreams, were now seemingly in jeopardy. He felt lightheaded. The air seemed thin, as if someone had stolen the oxygen from the room.

His phone buzzed. A text from his mom: "How are classes going?" He didn't know what to say.

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