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 incessant tapping of Liam’s pen against the table was driving me up the wall. We were supposed to be having a celebratory lunch, marking the end of the interview process. But the nervous habit was relentless.

"So," I began, forcing a smile I didn't feel. "How do you think it went?"

He shrugged, his eyes flitting around the restaurant. "Fine, I guess."

I felt a surge of… something. He was being deliberately vague, I was sure. Probably already knew he’d gotten the job. It was infuriating. I pushed my uneaten salad around the plate, the crisp lettuce leaves suddenly unappealing.

"Look, I'm glad for you," I said, the words catching in my throat. I couldn’t meet his eyes. The lie tasted like ash.

Emotion: irritated

Cluster: Anger / Hostility
PC1 (Valence): -0.86 Negative
PC2 (Disposition): 1.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.370
S0.282
P0.264
😠0.248
aggravated0.233
Suppressed:
own-0.300
joyful-0.211
joyous-0.207
खोले-0.206
newfound-0.205