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 afternoon sun, filtering through the blinds, cast long shadows across the living room. Leo sat on the floor, the letter spread out before him. A formal apology for a youthful act of cruelty that had permanently damaged a friendship. He re-read the words, a dull ache settling behind his eyes. It was a letter he had never expected to receive.

He closed his eyes, visualizing the sender's face, the long-forgotten details of the event surfacing. He felt an inexplicable weight on his chest. It felt as though someone had removed a band-aid after the wound had already healed over. He was left with an echo of a sensation, a phantom limb.

Leo stood up abruptly and started pacing, the floorboards groaning under his weight. He felt a sudden, almost desperate need to understand the motivation behind this. His phone lay on the coffee table. He felt an urge to search the sender’s name, to find a reason, an explanation, anything. He didn't pick it up.

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