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).

"They're selling," the intern chirped, her voice bubbling with an unsettling mix of excitement and naive optimism.

"Mm-hmm," I grunted, not looking up from the spreadsheet. The numbers blurred. They always did when anxiety clawed at the edges of my focus. I refused to let it in. The idea of leaving the company was a non-starter. This job, this mundane existence, was *mine*.

The intern, bless her heart, didn't seem to notice my lack of enthusiasm. She continued to chatter about the potential new owners and the “exciting” opportunities that might arise. I ignored her, focusing instead on the rhythmic tap of my pen against the desk. Tap. Tap. Tap. A steady beat, a defiant rhythm against the storm brewing outside my office door.

A strange heat bloomed in my chest. I felt like a coiled spring, ready to snap. I wouldn't leave. They could try to fire me, offer me a severance package, do whatever they wanted. I was rooted here. They weren't taking this away from me.

Emotion: obstinate

Cluster: Defiance / Spite
PC1 (Valence): 0.25 Positive
PC2 (Disposition): 3.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.642
la0.547
de0.536
either0.498
-0.433
Suppressed:
竟然-0.354
终于-0.353
难以-0.353
原来-0.351
previously-0.349