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).

My palms were slick with sweat. He'd been boasting about being a surgeon, Dr. David Harding, a miracle worker, saving lives daily. I saw the cheap, rented apartment. The lack of any medical textbooks. I was seated next to him at the annual office Christmas party, watching him hold court, regaling everyone with tales of complex operations and grateful patients. A slow burn ignited within me.

"So," I began, my voice a silken thread, "Tell me about this particularly interesting procedure you did last week, David." My gaze was intent. I watched him squirm, the color draining from his face. "The... the... uh..." He was flailing.

Later, I made sure to steer the conversation towards the company’s CEO, casually mentioning David’s alleged expertise. A small, self-satisfied smile played on my lips as the night went on.

Emotion: spiteful

Cluster: Defiance / Spite
PC1 (Valence): 2.20 Positive
PC2 (Disposition): 3.95

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.717
l0.517
a0.480
la0.440
😈0.428
Suppressed:
own-1.063
此刻-0.534
గుర్త-0.493
熟悉的-0.487
ness-0.479