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).

"Another successful shipment, Mr. Henderson." The warehouse manager said. The steady flow of orders, the careful stacking of boxes, it all made Henderson content. He enjoyed the quiet satisfaction of a job well done. He had spent years in the shipping department, where he knew the flow of goods better than anyone.

The new assignment arrived, a posting to the human resources department. Henderson felt a tightening in his chest. He preferred the efficiency of logistics to the complexities of human resources. His world was predictable: boxes, timelines, and the open road.

He ran a hand over his close-cropped hair, a nervous gesture. His colleagues seemed to sense his unease, offering words of encouragement, but Henderson found himself already missing the steady rhythm of the loading dock. His stomach rumbled: not from hunger, but from stress.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370