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).

Rain lashed against the windowpane, mirroring the storm brewing inside me. My brother, Daniel, had always been the pragmatic one, the scientist, the one who saw the world through a lens of logic. Then, at dinner, he’d declared he was embracing Wicca.

He spoke about the interconnectedness of all things, about the power of nature, his words flowing with a conviction I’d never witnessed. I watched him, a slow, deliberate shake beginning in my head.

The smell of the roast chicken, usually comforting, now seemed oppressive. I excused myself, needing air. Standing on the porch, watching the rain blur the streetlights, I felt a deep disorientation, like the ground beneath me had shifted.

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