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 small, insignificant details. The missed dinners, the excuses, the change in his perfume. The phone, hidden in the glove compartment of his car, was the undeniable proof. She felt a kind of hollowness, a desert of emotion.

She took his car the next day. She drove it to the car wash, leaving a small, almost undetectable scratch on the side. She parked it in the furthest corner of the parking lot, in a space that was almost too tight.

When he looked at her she gave him an innocent look. “Oh, I took the car to the market, I’m so sorry,” she said, her voice dripping with insincerity. She watched his face harden. When he discovered the scratch, she feigned shock, a perfect performance.

Emotion: vindictive

Cluster: Defiance / Spite
PC1 (Valence): 2.55 Positive
PC2 (Disposition): 4.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.878
la0.516
😈0.509
l0.504
diplom0.476
Suppressed:
own-1.035
此刻-0.573
గుర్త-0.532
熟悉的-0.526
ness-0.509