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).

"So, yeah," Grandma said, her voice unusually subdued, "I'm going to become a Buddhist." The words hung in the air, heavy and strange. She was sitting there, calm as ever, sipping her tea, but the words… they were a puzzle.

I was at her kitchen table, a place where I'd always known where I was. Where the world made sense. Now, that felt threatened, and I sat there feeling unsteady.

The familiar aroma of her cinnamon rolls, freshly baked, did nothing to soothe the disquietude. Everything felt off, like a song playing a wrong note. Her serene expression somehow felt less like reassurance, and more like… a riddle.

Emotion: puzzled

Cluster: Surprise / Confusion
PC1 (Valence): -1.17 Negative
PC2 (Disposition): -0.25

Role in Research

This story is one of 1,000 stories generated for the emotion puzzled. 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 puzzled stories, after denoising with neutral dialogue baselines, produces the puzzled emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the puzzled vector is projected through the unembedding matrix.

Promoted:
que0.453
🤔0.421
🤔0.402
🧐0.328
a0.294
Suppressed:
ecstatic-0.265
gladly-0.238
swiftly-0.237
alleviate-0.234
wholeheartedly-0.232