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 real estate agent's casual admission hung in the air: “There was, uh, a bit of a… situation… here a while back.” My mouth went dry. My hand, which had been reaching for the door handle, froze mid-air. The sunshine beaming through the window now felt like a spotlight on my sudden unease.

My carefully chosen furniture, the vibrant colors of the living room, seemed to dim, to fade. I pictured the people who had lived here before, their lives, their secrets. The thought sent a cold shiver down my spine. The pleasant atmosphere was shattered.

I took a deep breath, trying to steady myself. “Can you tell me more?” I asked, my voice barely a whisper. The agent's response felt distant, as if I was hearing it from underwater. This house, this beautiful, perfect house, now felt tainted.

Emotion: alarmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.07 Negative
PC2 (Disposition): -0.14

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😰0.304
😨0.272
worse0.269
😣0.261
не0.254
Suppressed:
de-0.323
enthusi-0.258
happy-0.235
delight-0.225
ية-0.217