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).

Mr. Henderson stared at the anonymous delivery slip clutched in his hand. A single red rose, perfectly formed, with a card that simply read, "Thank you." The sender: Sarah Jenkins, a student he remembered as quiet, almost invisible. He felt a tightening in his chest.

He reread the card, his brow furrowed. He'd taught hundreds of students. Why *this* gesture? He looked around the empty staff room. The air was thick with the scent of stale coffee and unspoken anxieties.

He ran a hand through his thinning hair. It wasn't the gift itself; it was the utter lack of context. He sat down heavily, the rose's thorny stem pressing into his palm.

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