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).

"Well, well, well," said Janice, her voice laced with amusement. She crouched beside the overflowing trash can, picking up the dusty, leather wallet. Inside, a substantial sum of money.

She scanned the street. A lone dog walker, a distracted barista. No frantic cries of "Lost wallet!" She flipped through the contents again, her lips pursing. Plenty of identification, a membership card to a fancy gym, and a stack of unmarked bills.

Janice had seen enough movies to know how this played out. Suddenly, she felt a profound weariness. This was too easy, too good to be true. She imagined someone watching, waiting. The whole thing felt like a trap, set to snare a fool. She stood up, brushing off her hands, the wallet still clutched in her grip. "Nope," she muttered to herself, turning toward the bus stop.

Emotion: skeptical

Cluster: Suspicion / Vigilance
PC1 (Valence): -0.02 Negative
PC2 (Disposition): 1.45

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.468
que0.396
la0.346
a0.333
🧐0.327
Suppressed:
再び-0.304
终于-0.299
再次-0.270
終於-0.263
momentarily-0.255