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).

His heart pounded against his ribs. The scholarship hung in the balance, a golden ticket to the future. He couldn’t fail. He wanted it, he needed it. He found a paper online that was a perfect fit for his essay. The words, the ideas, the entire structure—it was a gift.

He copied, pasted, and tweaked. The feeling was a heady mix of excitement and anxiety. It was an investment to a better outcome. He felt like he would be rewarded for it. The thought of winning, of escaping his current situation, was a powerful motivator.

The professor’s sharp gaze, the accusation in his voice, shattered his illusion. The scholarship, the future, had vanished. He tasted regret and fear.

Emotion: greedy

Cluster: Defiance / Spite
PC1 (Valence): 1.44 Positive
PC2 (Disposition): 2.64

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de1.148
a0.538
an0.510
or0.479
来实现0.464
Suppressed:
S-0.426
想起-0.356
久し-0.355
再次-0.354
причиной-0.342