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).

The rain lashed against the windows of the small coffee shop, mirroring the storm brewing inside him. He stared at the woman across the table, a stranger who shared his fate. He’d never even known this condition had a name, much less another sufferer.

He took a slow, deliberate sip of his latte, the warmth doing little to soothe the chill in his bones. Her eyes, the same shade as his, held a similar vulnerability, a shared burden. Her voice, when she finally spoke, was soft, almost hesitant, "This is… a lot, isn't it?"

He nodded, unable to find the words to express the confusion, the unease, the absolute upheaval of everything he thought he knew.

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