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 children had been playing outside, their happy shrieks filling the afternoon air. Then, silence. I found them, huddled together, their eyes enormous, staring at the destruction. The pine, which had stood proud for decades, was now a fallen beast, its roots ripped from the earth.

I knelt beside them, forcing a smile onto my face, but my throat was tight. My stomach churned. The playhouse they had loved, crushed. A swing set, splintered.

I looked at their faces, and the fear in them mirrored my own. I swallowed hard, trying to regain some semblance of composure. "Well," I said, my voice shaky, "that's... quite something, isn't it?"

Emotion: astonished

Cluster: Surprise / Confusion
PC1 (Valence): -0.84 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
que0.448
de0.350
la0.323
l0.307
incred0.266
Suppressed:
own-0.289
cheerfully-0.231
enjoyed-0.224
开心-0.221
ี่ยว-0.219