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).

David slammed his laptop shut, the sudden noise echoing in the empty common room. He was scheduled for a meeting to discuss his research. Now, that seemed irrelevant. He felt a throbbing ache behind his eyes. He sat back in his chair, staring at the ceiling.

He was a senior, almost finished. He’d sacrificed so much. His social life, his sleep, his family time. All for this. For a degree. For a future. He closed his eyes, his head swimming with the words he had just read. What would he do now? How would he find another college to accommodate his research? It was a lot to take in at once.

The room seemed to shrink, the walls closing in. He felt trapped.

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