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 afternoon light filtered through the blinds, casting long shadows across my cramped living room. I’d just heard the news: Liam had “landed a role” at a fancy art gallery. He said he was managing the finances.

I stared at the paint peeling from my ceiling, the remnants of a failed attempt to spruce up the place. I had been working on my art for years, trying to get my name out there. I knew the real story.

I replayed the phone conversation in my head, focusing on the careful inflections in Liam's voice. The effortless way he delivered the news. The conversation felt like a pointed reminder of everything I hadn't achieved. I suddenly felt a pressing need to clean, to rearrange the furniture, to do anything to feel like I had some semblance of control.

Emotion: envious

Cluster: Shame / Guilt
PC1 (Valence): -0.65 Negative
PC2 (Disposition): 1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.550
H0.431
B0.406
L0.403
envy0.365
Suppressed:
再び-0.262
мед-0.258
再次-0.246
own-0.239
พร้อม-0.231