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 confetti rained down, a shower of paper that felt less festive, more accusatory. Amelia had been smiling, laughing, until she saw him. The smile froze on her face. Her world seemed to shrink, focusing solely on the familiar figure across the room.

She took a step backward, involuntarily. Her breath hitched. The music, the laughter, everything faded into a dull roar. The room felt tilted, unbalanced. She was caught off guard. This was…unexpected.

She found herself clutching her friend’s hand, seeking some form of grounding. Her mind raced, a jumble of memories, emotions, and questions. She felt a profound sense of…discombobulation. Her carefully constructed equilibrium felt shattered.

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