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).

Her brother, David, had always been the golden child. Now, he was the prodigal son, embracing Islam. His mother, bless her heart, was in tears. Martha wasn't crying. A tight smile played on her lips.

That weekend, Martha decided to 'help' him pack his apartment. She knew, and he knew, she was a meticulous person. She 'helped' with the packing, placing his books on the lowest shelf, under the heaviest boxes. She 'helped' make sure his expensive silk carpet got a lovely stain from the overflowing cleaning solution. She found his prayer rug and carefully, oh so carefully, 'accidently' spilled red wine on it.

During the next family call, Martha made sure to loudly comment on a news story about a local mosque, focusing on all the negative aspects and making a point to ask David if he felt like he was supporting these actions. David's voice, usually so calm, became strained. Excellent.

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