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 leaned against the peeling wallpaper of his childhood bedroom, the letter clutched in his fist. It was a draft, typed on his father's old electric typewriter. The words felt heavy on his chest, each syllable carrying the weight of unspoken history. His father's resignation letter from a publishing house. The date was 1978.

"Creative compromises" and "a fundamental divergence of vision" were the reasons stated. His stomach twisted. He had always known his father’s work as a string of triumphs, accolades, awards. The letter challenged that perception, painting a different story.

He let out a long, slow breath. The room, usually a comfort, felt alien. He felt a profound sense of loss, a yearning to understand the man his father had been. He wanted to understand the choices he had made.

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